Explore Your Local Site

Looks like you've landed on our   site. Let's take you home:    

Please note that the content and products on the    site might not be available in your region.

 

Choose the language:

  Homepage
Continue on the current website:  

 

Miglioramento dello screening oncologico: in che modo l'IA sostiene il lavoro dei radiologi

Il cancro è una delle principali cause di morbilità e mortalità a livello globale: in media una persona su cinque sviluppa un tumore nel corso della vita (The Burden of Cancer, senza data). Lo screening oncologico è rivolto a individui asintomatici e mira a identificare tumori in stadio iniziale o lesioni precancerose. In molti casi, ciò consente di intervenire tempestivamente e di ottenere esiti migliori. Lo screening costituisce dunque sostanzialmente uno strumento per la prevenzione e la diagnosi precoce. Lo screening preventivo mira a individuare patologie benigne che possono evolvere in forme maligne, evento possibile solo nel caso di alcuni tumori, mentre gli approcci per la diagnosi precoce sono finalizzati a individuare il tumore nella fase iniziale. È importante sottolineare che lo screening non deve essere considerato un unico test, bensì un percorso che include l’identificazione della popolazione target, l’esecuzione di test diagnostici e la pianificazione di ulteriori accertamenti, incluso il trattamento laddove necessario (Organizzazione Mondiale della Sanità. Ufficio regionale per l’Europa, 2022).

La radiologia svolge da tempo un ruolo fondamentale nello stabilire l’entità della diffusione locale e a distanza del tumore dopo la diagnosi, tuttavia è indispensabile anche nei percorsi di screening di diversi tumori comuni. In tali casi, gli studi di imaging medico rappresentano lo strumento di screening primario o vengono utilizzati per prendere decisioni in merito all’esecuzione di ulteriori accertamenti dopo lo screening utilizzando altre metodiche, tra cui le analisi del sangue. A seconda del tipo di tumore, lo screening può prevedere tecniche di imaging medico tra cui mammografia, tomografia computerizzata (TC), risonanza magnetica (RM) ed ecografia. Sono stati istituiti programmi nazionali di screening che utilizzano l’imaging medico per alcuni dei tumori più comuni. La maggior parte di questi programmi è rivolta a specifiche popolazioni a rischio di sviluppare il tumore specifico oggetto di indagine, identificate sulla base di fattori di rischio modificabili o non modificabili.

Poiché lo screening oncologico è pensato per soggetti sani, è particolarmente importante che i vantaggi di un programma di screening siano superiori agli svantaggi. Ciò deve essere stabilito attentamente per ciascun programma ed è talvolta controverso (Lam et al., 2014). Tuttavia, alcuni pro e contro dello screening oncologico valgono per tutte le tecniche di screening e per tutti i tumori (Kramer, 2004; Organizzazione Mondiale della Sanità. Ufficio regionale per l’Europa, 2022). Lo screening può ridurre la spesa sanitaria e migliorare la qualità della vita dei pazienti. Spesso migliora anche la prognosi e gli esiti del trattamento per gli individui identificati come affetti da tumore e può essere rassicurante per gli individui che ottengono risultati negativi. Tuttavia, talvolta la diagnosi precoce non cambia la prognosi e, in questi soggetti, lo screening può indurre a trattamenti non necessari, con conseguente scadimento della salute o della qualità della vita. Infatti, talvolta lo screening può rilevare tumori che non comprometteranno mai la salute nel corso della vita né causeranno la morte dell’individuo interessato. Inoltre, tutti i test di screening producono inevitabilmente falsi positivi e falsi negativi, determinando nel primo caso sovratrattamento, con i conseguenti effetti indesiderati a livello psicosociale e fisico, e nel secondo caso false rassicurazioni e ritardo nel trattamento.

Carcinoma mammario

Il carcinoma mammario è la seconda causa di morte per cancro nelle donne (Bray et al., 2018) ed è uno dei tumori più comuni a livello mondiale (Sung et al., 2021). La diagnosi e il trattamento precoci possono portare a esiti migliori, e gli studi condotti hanno dimostrato una riduzione della mortalità fino al 20% nelle popolazioni sottoposte a screening rispetto a quelle a cui non è stato offerto lo screening. Secondo le stime, si evita in media una morte per carcinoma mammario ogni 250-414 donne sottoposte a screening (Marmot et al., 2013; Tabár et al., 2011). Oltre 100 Paesi in tutto il mondo hanno implementato programmi di screening su larga scala per il carcinoma mammario (Existence of National Screening Program for Breast Cancer, senza data). Si raccomanda di iniziare lo screening tra i 40 e i 50 anni (Ren et al., 2022) e di eseguirlo quasi esclusivamente tramite mammografia, che utilizza una bassa dose di raggi X per l’acquisizione di immagini della struttura della mammella, o tramite tomosintesi mammaria digitale, una tecnica simile che utilizza proiezioni multiple per ottenere una serie di immagini che poi, sovrapposte, ricostruiscono la struttura della mammella.

 

breast cancer ebook

L’algoritmo ha aumentato del 12-27% il rilevamento del carcinoma mammario selezionando le mammografie refertate come negative dopo la doppia lettura, ma considerate sospette dall’algoritmo, per un’ulteriore valutazione mediante RM o ecografia.

L'accuratezza della mammografia varia considerevolmente, e anche la lettura dei radiologi più esperti produce tassi elevati di falsi positivi e falsi negativi (Elmore et al., 2009; Lehman et al., 2015). Si stima che almeno una donna su tre sottoposte a screening avrà un risultato falso positivo alla mammografia nel corso della vita (Castells et al., 2006). La lettura della mammografia è particolarmente impegnativa nelle donne con seno denso (Boyd et al., 2007) e in quelle sottoposte a terapia ormonale sostitutiva (Banks et al., 2006). Inoltre, lo screening mammografico è un’attività che richiede molto lavoro. In molti Paesi europei, lo standard di cura è la doppia lettura con discussione consensuale, in cui ogni caso è valutato consecutivamente da due radiologi che risolvono le divergenze consensualmente (Giordano et al.,2012). Purtroppo, in molti Paesi vi è anche una carenza di radiologi e di tecnici di radiologia specificatamente formati in mammografia (Moran & Warren-Forward, 2012; Rimmer, 2017; Wing & Langelier, 2009).

In varie fasi del processo di screening per carcinoma mammario sono stati incorporati sistemi basati sull’intelligenza artificiale (IA). Uno studio condotto negli Stati Uniti e nel Regno Unito su quasi 30.000 donne sottoposte a mammografie di screening a intervalli di 1-3 anni e a un follow-up esteso fino a 39 mesi ha confrontato una combinazione di tre modelli di deep learning (apprendimento profondo) con il referto istologico e le interpretazioni di radiologi abilitati (McKinney et al., 2020). L’algoritmo ha mostrato una specificità e una sensibilità più elevata, rispettivamente dell’1,2-5,7% e del 2,7-9,4%, rispetto ai radiologi che hanno eseguito la prima lettura. Secondo gli autori, l’utilizzo dell’algoritmo potrebbe rendere superflua la seconda lettura addirittura nell’88% dei casi di screening, a parità di accuratezza, liberando in tal modo risorse di cui si avverte estrema necessità.

Risultati altrettanto promettenti sono stati riferiti in studi con l’utilizzo di sistemi basati sull’IA a fianco dei radiologi. Secondo uno studio condotto in Spagna su circa 16.000 donne sottoposte a mammografia digitale o tomosintesi mammaria digitale, l’uso di un algoritmo di deep learning ridurrebbe il carico di lavoro del 72,5% rispetto alla doppia lettura, mantenendo invariata la sensibilità (Raya-Povedano et al., 2021). In questo modello, gli esami meno sospetti sarebbero letti solo dall'algoritmo e il 2% degli esami valutati come altamente sospetti dall’algoritmo sarebbe segnalato per l’esecuzione di ulteriori accertamenti, indipendentemente dall'interpretazione dei radiologi. Parimenti, uno studio condotto in Svezia su 7.354 donne ha evidenziato che un algoritmo di deep learning disponibile in commercio ha classificato accuratamente le mammografie meno sospette di alcune donne, che sono state sottoposte a ulteriori accertamenti (Dembrower et al., 2020). Questi risultati sono stati ottenuti con un tasso di falsi negativi dello 0-2,6%. L'algoritmo ha inoltre aumentato del 12-27% il rilevamento del carcinoma mammario selezionando le mammografie refertate come negative dopo la doppia lettura, ma considerate sospette dall'algoritmo, per un'ulteriore valutazione mediante RM o ecografia.

Altri studi hanno utilizzato sistemi basati sull’IA nel processo volto a determinare se fossero necessari ulteriori accertamenti (decision referral). In uno studio condotto in Germania su oltre un milione di mammografie, una rete neurale convoluzionale (Convolutional Neural Network, CNN) profonda ha assegnato un punteggio di confidenza a ciascuna mammografia (Leibig et al., 2022). Per le valutazioni effettuate dall'algoritmo con elevata confidenza non sono state richieste ulteriori indagini, mentre le valutazioni con bassa confidenza sono state inviate all’attenzione del radiologo. Questo approccio è stato associato a un aumento del 4% della sensibilità e dello 0,5% della specificità rispetto alla valutazione di un solo radiologo senza l'ausilio dell'algoritmo. In questo scenario, il 63% delle mammografie è stato sottoposto automaticamente a triage dall'algoritmo, e il miglioramento delle prestazioni rispetto alla lettura di un solo radiologo è stato coerente in otto centri di screening e per tre marche di dispositivi.

Carcinoma polmonare

Il carcinoma polmonare è la principale causa di morte per cancro in tutto il mondo, con quasi 1,8 milioni di decessi nel 2020 (Sung et al., 2021). Lo screening per il carcinoma polmonare rappresenta un esempio di approccio mirato ed è raccomandato in funzione del rischio individuale. Lo screening degli adulti di età compresa tra 50 e 80 anni che abbiano fumato 20 pacchetti all’anno mediante tomografia computerizzata a basso dosaggio (Low-Dose Computed Tomography, LDCT) è raccomandato negli USA dal 2013, dopo che gli studi iniziali hanno mostrato una riduzione relativa della mortalità per carcinoma polmonare del 20% (Lung Cancer: Screening, 2021; National Lung Screening Trial Research Team et al., 2011). Un analogo programma di screening è in fase di implementazione nel Regno Unito (NHS England, 2022).

lung cancer ebook

Dallo studio è emerso che l’uso dell’algoritmo è associato a una migliore sensibilità per il rilevamento dei noduli, indipendentemente dal livello di esperienza del primo lettore.

Nei pazienti sottoposti a screening per il carcinoma polmonare mediante LDCT, l’IA si è dimostrata promettente nella rilevazione automatica di noduli polmonari potenzialmente maligni. Ciò è importante perché la rilevazione dei noduli polmonari da parte dei radiologi è impegnativa, dispendiosa in termini di tempo e soggetta a errori (Al Mohammad et al., 2019; Armato et al., 2009; Gierada et al., 2017; Leader et al., 2005). Uno studio condotto su quasi duemila pazienti ha testato come secondo lettore un algoritmo basato su una CNN concepito per rilevare automaticamente i noduli polmonari (Katase et al., 2022). La verità di base era costituita da noduli identificati da due radiologi esperti come ad alto rischio in base all’anamnesi clinica e alle caratteristiche morfologiche del nodulo. Dallo studio è emerso che l'uso dell'algoritmo è associato a una migliore sensibilità per il rilevamento dei noduli, indipendentemente dal livello di esperienza del primo lettore. Sebbene la sensibilità complessiva sia risultata più bassa per i noduli a vetro smerigliato e per i noduli di diametro inferiore a 1 cm, la sensibilità per questi ultimi è risultata molto più elevata quando è stato utilizzato l'algoritmo rispetto a quando l’interpretazione delle immagini è stata affidata al radiologo da solo. I falsi positivi hanno incluso aree di infiammazione pleurica o vasi periferici, mentre i falsi negativi sono stati rappresentati spesso da noduli a vetro smerigliato debolmente o scarsamente delimitati o da noduli prossimi al diaframma. Si fa presente che gli autori hanno riscontrato prestazioni del modello coerenti con varie dosi di radiazioni della TC nell’ambito di uno studio su fantoccio, a indicare che i loro risultati potrebbero essere generalizzabili ad altri protocolli di TC del torace (Katase et al., 2022). Un altro studio ha rilevato una sensibilità del 93% e una specificità del 96% per un algoritmo basato su una CNN per la rilevazione di noduli polmonari alla LDCT rispetto al consenso di due radiologi (Chamberlin et al., 2021). In tale studio, i falsi positivi hanno incluso aree di atelettasia, alterazioni parenchimali associate a infezione e osteofiti protrudenti dalle vertebre toraciche che interferivano con i campi polmonari.

Oltre alla pura e semplice identificazione dei noduli polmonari, alcuni studi hanno tentato di classificare il rischio di malignità dei noduli identificati. Un algoritmo multicomponente comprendente segmentazione del volume polmonare, identificazione dell’area interessata dal tumore e modelli di previsione del rischio di tumore è stato testato su 6.716 LDCT e validato su un set indipendente di dati ottenuti da 1.139 LDCT (Ardila et al., 2019). L'algoritmo genera una probabilità di malignità in funzione di singole LDCT o, se disponibili, di precedenti LDCT dello stesso paziente. Utilizzando come verità di base casi di carcinoma polmonare confermati mediante biopsia, l'algoritmo ha fornito le stesse prestazioni di sei radiologi quando erano disponibili referti di LDCT precedenti. Nei casi in cui non erano disponibili precedenti LDCT, l'algoritmo ha prodotto tassi di falsi positivi e falsi negativi più bassi (rispettivamente 11% e 5%) rispetto ai radiologi.

La valutazione del parenchima polmonare alla LDCT che vada oltre il solo rilevamento di eventuali noduli polmonari rappresenta un approccio recente e promettente per identificare il rischio di sviluppare in futuro un carcinoma polmonare. Uno studio ha rilevato che un algoritmo CNN 3D, testato su oltre 15.000 LDCT, aveva un’area sotto la curva (Area Under the Curve, AUC) della caratteristica di funzionamento del ricevitore (ROC) di 0,86-0,94 (a seconda del set di dati) per la previsione dell’insorgenza di un carcinoma polmonare a un anno (Mikhael et al., 2023). È interessante notare che l'AUC dell'algoritmo dopo aver escluso i casi con presenza di noduli visibili al basale nella stessa posizione dei futuri tumori è stata pari a 0,82. L'algoritmo ha inoltre mostrato un tasso di falsi positivi più basso rispetto a quello dei punteggi di malignità stabiliti, sulla base delle caratteristiche morfologiche del nodulo, quando l'intero volume di LDCT è stato valutato dall'algoritmo. Da questi risultati si evince che altre caratteristiche oltre ai noduli sospetti hanno contribuito alla previsione generata dall’algoritmo. Soprattutto, ciò significa che l’algoritmo rileva nella LDCT altre caratteristiche oltre a quelle che i radiologi considerano in genere rilevanti per prevedere il rischio di sviluppare una neoplasia polmonare.

I criteri di idoneità allo screening per il carcinoma polmonare negli Stati Uniti, dettati dai Centers for Medicare and Medicaid Services (CMS), non rilevano oltre la metà dei casi di carcinoma polmonare (Y. Wang et al., 2015). Sebbene esistano altri strumenti di “pre-screening” più complessi e basati su punteggi, le informazioni che richiedono, come il numero di pacchetti-anno, sono spesso imprecise o non disponibili (Kinsinger et al., 2017). L’IA è stata quindi utilizzata per identificare un maggior numero di individui ad alto rischio per il carcinoma polmonare al fine di inserirli nei programmi di screening. Uno studio condotto su 5.615 soggetti ha rilevato che una combinazione di fattori quali radiografie semplici del torace, età, sesso e attuale stato di tabagismo consente una selezione più mirata dei pazienti per lo screening mediante LDCT (Lu et al., 2020). Il modello utilizzato in questo studio ha dato un'AUC di 0,7 per la previsione del rischio di insorgenza di un carcinoma polmonare incidente a 12 anni rispetto a un'AUC di 0,63 per i criteri dei CMS, il che si traduce in una riduzione del 30,7% dei casi di carcinoma polmonare non diagnosticato con l'algoritmo. Il modello ha previsto anche una mortalità per carcinoma polmonare a 12 anni con un’AUC di 0,76. Gli autori non raccomandano radiografie del torace di routine per il pre-screening, ma sostengono l'uso di questo modello nei pazienti sottoposti a radiografie del torace per altre indicazioni cliniche.

Carcinoma del colon-retto

Il carcinoma del colon-retto è il terzo tumore più comune sia nelle donne sia negli uomini ed è una delle principali cause di morte per cancro in tutto il mondo (Sung et al., 2021). Si sviluppa come una cascata di eventi, poiché le cellule della mucosa intestinale passano attraverso una serie di mutazioni genetiche, che la trasformano prima in mucosa iperproliferativa, poi in adenoma benigno e, in alcuni casi, in adenocarcinoma (Kuipers et al., 2015). Lo screening per il carcinoma del colonretto è principalmente preventivo. Mira infatti a rilevare adenomi potenzialmente maligni in modo che possano essere rimossi, un approccio che riduce la mortalità della malattia (Zauber et al., 2012).

Colorectal cancer ebook

Un recente studio proof-of-concept ha utilizzato un approccio completamente automatizzato basato sulle CNN per la segmentazione dei polipi e la distinzione tra polipi benigni e polipi precancerosi.

Lo screening per il carcinoma del colon-retto si esegue di routine tramite la ricerca di sangue occulto nelle feci utilizzando test altamente sensibili oppure tramite lo studio del lume intestinale mediante colonscopia ottica (Helsingen Lise M. & Kalager Mette, 2022). La colonscopia ottica è un metodo consolidato e affidabile per l’identificazione degli adenomi del colon-retto, che possono essere rimossi contestualmente. Tuttavia, i suoi principali svantaggi sono la scarsa compliance del paziente e la necessità di sedazione (Inadomi et al., 2012; Joseph et al., 2012; OECD, 2012; Stock et al., 2011; Use of Colorectal Cancer Screening Tests, 2023).

Una promettente alternativa emergente alla colonscopia ottica è la colonografia con tomografia computerizzata (CTC), una tecnica che ha un’accuratezza diagnostica simile alla colonscopia ottica (Pickhardt et al., 2003, 2011, 2018), è preferita dai pazienti (Ristvedt et al., 2003) ed è associata a una migliore compliance (Moawad et al., 2010). Inoltre, non richiede sedazione ed è in grado di rilevare reperti clinicamente rilevanti in sede extra-intestinale che sono invisibili alla colonscopia ottica (Smyth et al., 2013). Tuttavia, la CTC richiede una preparazione intestinale (come la colonscopia ottica), espone il paziente a radiazioni ionizzanti e non consente la resezione simultanea di eventuali polipi. Nonostante questi svantaggi, l’American College of Radiology raccomanda la CTC per lo screening dei pazienti con rischio medio o moderato per carcinoma del colon-retto (Expert Panel on Gastrointestinal Imaging: et al., 2018).

Le immagini della CTC passano attraverso una serie di passaggi preliminari prima di essere interpretate, comprendenti la pre-elaborazione per l’eliminazione degli artefatti, la separazione del colon dalle restanti strutture addominali, la ostruzione 3D del colon e la visualizzazione del lume colico. Uno studio recente ha combinato un nuovo metodo di segmentazione e ricostruzione del colon con la rilevazione di polipi mediante una CNN (Alkabbany et al., 2022). La segmentazione automatizzata del colon è risultata sovrapponibile per oltre il 90% con la segmentazione manuale effettuata da un esperto nel 70% dei casi, e i polipi del colon sono stati rilevati con un'AUC di 0,93, una sensibilità del 97% e una specificità del 79%.

La differenziazione tra polipi benigni e polipi potenzialmente maligni è una sfida sia alla colonscopia ottica sia alla CTC ed è stato l’argomento di numerosi studi basati sull’IA. Gli approcci basati sulla radiomica per la classificazione dei polipi benigni rispetto a quelli precancerosi alla CTC hanno mostrato AUC fino a 0,91, ma richiedono la segmentazione manuale dei polipi (Grosu et al., 2021; Song et al., 2014). Un recente studio proof-of-concept ha utilizzato un approccio completamente automatizzato basato sulle CNN per la segmentazione dei polipi e la distinzione tra polipi benigni e polipi precancerosi (Wesp et al., 2022). Gli autori hanno addestrato la CNN con dati di 63 pazienti e l’hanno testata su un set di dati indipendente relativo a 59 pazienti, mostrando un'AUC fino a 0,83 e una sensibilità e specificità dell'80% e del 69%, rispettivamente. Tali approcci basati sull’IA possono potenzialmente essere utilizzati come secondo lettore per guidare la decisione relativa alla rimozione del polipo.

Carcinoma epatocellulare

Il carcinoma epatocellulare (Hepatocellular Carcinoma, HCC) è una delle più comuni cause di morte per cancro nel mondo (Sung et al., 2021). Gli individui con cirrosi epatica o infezione cronica da virus dell'epatite B o C sono ad alto rischio di sviluppare un HCC (Vogel et al., 2022). Lo screening di questi pazienti è associato a una riduzione della mortalità per HCC (Singal et al., 2022; Zhang et al., 2004). Lo screening si esegue solitamente mediante ecografia addominale ogni sei mesi (European Association for the Study of the Liver, 2018; Frenette et al., 2019; Marrero et al., 2018), con o senza misurazione dei livelli ematici di alfa-fetoproteina (Colli et al., 2006; Tzartzeva et al., 2018). Le lesioni sospette identificate all'ecografia vengono ulteriormente caratterizzate mediante TC o RM o entrambe.

Hepatocellular cancer ebook

Le tecniche di deep learning sono state ampiamente applicate anche nell’imaging epatico utilizzando l’ecografia in modalità B e hanno mostrato risultati promettenti nel rilevamento delle lesioni epatiche focali e nella loro classificazione in benigne o maligne.

La patogenesi dell'HCC configura una complessa interazione tra noduli epatici presenti in diversi stadi del danno epatico cronico. I noduli rigenerativi si formano in risposta al danno a carico degli epatociti e si osservano comunemente in un fegato cirrotico. In questi noduli rigenerativi, nel tempo possono accumularsi mutazioni genetiche, che li convertono in noduli displastici ad alto rischio di progressione verso l’HCC man mano che si accumulano altre mutazioni (Kudo, 2009). Differenziare tra noduli displastici e noduli maligni mediante le tecniche di imaging è difficile (Park et al., 2017). Inoltre, le caratteristiche che l'HCC rivela all’imaging talvolta si sovrappongono a quelle di altre lesioni epatiche, inclusi emangiomi, cisti epatiche semplici e iperplasia nodulare focale (Heiken, 2007).

Utilizzando un approccio radiomico basato sulla combinazione dei dati relativi alla perfusione e dell’analisi della struttura ottenuta all’ecografia con mezzo di contrasto, uno studio condotto su 72 pazienti ha registrato un'accuratezza bilanciata di 0,84 nella differenziazione tra lesioni epatiche benigne e maligne (Turco et al., 2022). Un altro studio che ha utilizzato l’ecografia con contrasto ha riscontrato una sensibilità del 94,8% e una specificità del 93,6% nella differenziazione tra HCC e iperplasia nodulare focale utilizzando un approccio di apprendimento automatico basato sul metodo dei vettori di supporto (Huang et al., 2020); risultati comparabili sono stati ottenuti da altri studi (Gatos et al., 2015; Kondo et al., 2017). In uno studio multicentrico che ha valutato la differenziazione di 11 diversi tipi di lesioni epatiche focali utilizzando l'ecografia con mezzo di contrasto e l'istopatologia come riferimento, l’approccio di apprendimento automatico basato sul metodo dei vettori di supporto (AUC = 0,883) ha dato risultati di gran lunga migliori rispetto a quelli ottenuti con una rete neurale artificiale (AUC = 0,829), ed entrambi gli approcci si sono dimostrati superiori rispetto ai risultati ottenuti da un radiologo esperto (AUC = 0,702) (Ta et al., 2018).

Le tecniche di deep learning sono state ampiamente applicate anche nell’imaging del fegato utilizzando l’ecografia in modalità B. Questi studi hanno evidenziato risultati promettenti per la rilevazione di lesioni epatiche focali (Brehar et al., 2020; Schmauch et al., 2019; Tiyarattanachai et al., 2022) e la loro classificazione in benigne o maligne (Schmauch et al., 2019) oppure in entità specifiche (Hassan et al., 2017; Virmani et al., 2014). Utilizzando un approccio di deep learning, uno studio ha riscontrato che combinando i dati demografici del paziente e i risultati di laboratorio con le immagini dell’ecografia in modalità B, l’AUC per la classificazione delle lesioni epatiche in benigne vs. maligne è migliorata da 0,721 (solo ecografia) a 0,994 (Sato et al., 2022). Un altro studio condotto su 334 pazienti ha rilevato che il tasso di rilevazione delle lesioni epatiche focali all'ecografia in modalità B utilizzando una CNN è risultato più elevato per l'HCC rispetto ad altre lesioni epatiche focali, e la CNN si è imposta rispetto agli esperti umani (con un tasso di rilevazione dell'algoritmo del 100% rispetto al 39,1% per i non radiologi e al 69,6 % per i radiologi) (Tiyarattanachai et al., 2022).

Carcinoma prostatico

Il carcinoma prostatico è il tumore più frequente negli uomini in Europa e negli Stati Uniti (Ferlay et al., 2018; Siegel et al., 2021) ed è il terzo tumore più comune al mondo (Sung et al., 2021). Nei Paesi in cui esistono programmi di screening, questo si basa solitamente sulla misurazione dei livelli sierici dell’antigene prostatico specifico (Prostatespecific Antigen, PSA). Il PSA sierico ha un'elevata sensibilità ma una bassa specificità per il carcinoma prostatico (Merriel et al., 2022). Lo screening basato esclusivamente sul PSA porta quindi a molte biopsie non necessarie, con un tasso anche del 75% di biopsie prostatiche sistematiche (quelle in cui non si mira a un punto specifico all'interno della ghiandola, ma si ottengono più frammenti di tessuto da più parti della ghiandola) negative (Ahmed et al., 2017). Inoltre, lo screening basato sul PSA tende a rilevare tumori a basso rischio e a crescita più lenta che sono considerati clinicamente non significativi, perché non mettono in pericolo la sopravvivenza del paziente (US Preventive Services Task Force et al., 2018; Welch & Albertsen, 2020). Nel complesso, quindi, il vantaggio dello screening basato sui livelli sierici di PSA seguito da una biopsia sistematica è discutibile. Al contrario, l’approccio ideale è quello che rileva il tumore e contemporaneamente ne caratterizza la significatività clinica.

prostate cancer ebook

In uno studio, l’utilizzo di un classificatore Random Forest per il rilevamento di aree sospette alla RM multiparametrica della prostata è stato associato a tempi di lettura più brevi e una migliore specificità.

La RM multiparametrica svolge un ruolo sempre più importante nell'iter diagnostico dei casi screenati di carcinoma prostatico e include sequenze pesate in diffusione e in T2, con o senza sequenza dinamica post-contrasto pesata in T1 (Walker et al., 2020). L’uso della RM può ridurre i falsi positivi e la rilevazione di un carcinoma prostatico clinicamente non significativo e ciò può a sua volta contribuire a ridurre il sovratrattamento (Drost et al., 2019). Secondo alcuni studi, la RM prima della biopsia può ridurre di un terzo il numero di biopsie non necessarie (Elwenspoek et al., 2019), e quest’approccio è stato incluso in diverse linee guida sul trattamento del carcinoma prostatico ( Leitlinienprogramm Onkologie: Prostatakarzinom, senza data, Overview | Prostate Cancer: Diagnosis and Management | Guidance | NICE, senza data; Mottet et al., 2017). La RM può anche contribuire a eseguire biopsie mirate in pazienti con biopsie prostatiche sistematiche negative (Hoeks et al., 2012; Hugosson et al., 2022; Penzkofer et al., 2015; Siddiqui et al., 2015; Sonn et al., 2014). Nei pazienti affetti da carcinoma prostatico a rischio molto basso o basso, la RM può essere utile per il monitoraggio attivo della malattia, un approccio associato a buoni esiti a lungo termine (Klotz et al., 2015). La lettura delle RM della prostata è, tuttavia, difficile e anche i sistemi di refertazione standardizzati hanno una curva di apprendimento ripida, con prestazioni diagnostiche notevolmente diverse da radiologo all’altro e da una struttura all’altra (Kohestani et al., 2019; Muller et al., 2015; Rosenkrantz et al., 2017; Smith et al., 2019; Westphalen et al., 2020).

La segmentazione dell'intera prostata consente di stabilire il volume della ghiandola, utile per il calcolo della densità del PSA (una metrica che aiuta a differenziare tra ipertrofia prostatica benigna e carcinoma prostatico) e per la pianificazione della radioterapia. Tuttavia, la segmentazione manuale della prostata da parte dei radiologi richiede molto tempo ed è soggetta a errori (Garvey et al., 2014). La segmentazione automatizzata della prostata mediante strumenti basati sull'IA è fattibile e accurata, e diversi strumenti attualmente disponibili in commercio sono adatti per questo scopo (AI for Radiology, n.d.; Bardis et al., 2021; Belue & Turkbey, 2022; Sanford et al., 2020; Sunoqrot et al., 2022; Turkbey & Haider, 2022; Ushinsky et al., 2021; van Leeuwen et al., 2021; B. Wang et al., 2019).

Gli approcci basati sull’IA si sono rivelati utili anche per l’identificazione e la segmentazione del carcinoma prostatico alla RM multiparametrica. Gli algoritmi generalmente classificano le lesioni in due classi (ad esempio, carcinoma prostatico clinicamente significativo vs. carcinoma prostatico clinicamente non significativo) o in più classi utilizzando il punteggio PI-RADS (Belue & Turkbey, 2022; Twilt et al., 2021). In uno studio multi-lettore e multicentrico, l’utilizzo di un classificatore Random Forest per il rilevamento di aree sospette alla RM multiparametrica della prostata è stato associato a tempi di lettura più brevi (2,7-4,4 minuti con l’algoritmo rispetto a 3,5-6,3 minuti senza l’algoritmo, a seconda dell’esperienza del lettore) e a una migliore specificità (71,5% vs. 44,8%) (Gaur et al., 2018).

Diversi studi in cui sono stati utilizzati approcci di deep learning hanno raggiunto AUC fino a 0,89 per la rilevazione del carcinoma prostatico alla RM multiparametrica (Arif et al., 2020; Saha et al., 2021). Un algoritmo di deep learning disponibile in commercio ha migliorato la rilevazione del carcinoma prostatico clinicamente significativo da parte dei radiologi (utilizzando il consenso di tre radiologi esperti come riferimento), ha aumentato l'affidabilità inter-reader e ha ridotto il tempo di lettura mediano (Winkel et al., 2021). Come per il carcinoma mammario, l’accuratezza diagnostica è massima quando gli strumenti basati sull’IA e l’interpretazione del radiologo vengono utilizzati congiuntamente, anziché fare affidamento sulla valutazione esclusivamente dell’uno o dell’altro (Cacciamani et al., 2023).

L’IA è stata utilizzata anche per classificare l’aggressività del carcinoma prostatico. In uno studio di radiomica basato sulla RM, è stato utilizzato un classificatore basato su una macchina a vettori di supporto per la segmentazione delle aree di carcinoma prostatico, seguita da analisi della struttura ed estrazione delle caratteristiche quantitative (Giannini et al., 2021). Nello stesso studio, un altro classificatore basato su una macchina a vettori di supporto ha utilizzato le caratteristiche estratte per classificare l'aggressività del tumore utilizzando la classificazione istopatologica come riferimento. Lo studio, che prevedeva l’addestramento con dati di 72 pazienti, ha rilevato un'AUC di 0,81 in un set di dati di convalida di 59 pazienti (valore predittivo positivo = 81%, valore predittivo negativo = 71%). In un altro studio su RM multiparametriche della prostata di 107 pazienti, la classificazione PI-RADS secondo il radiologo è stata combinata con un punteggio di probabilità ottenuto con un classificatore Random Forest e tutte le regioni sospette identificate in questo modo sono state sottoposte a biopsia (Litjens et al., 2015). L'inclusione del punteggio dell'algoritmo è stata associata a una maggiore probabilità di rilevamento del carcinoma prostatico (AUC = 0,88 con l’algoritmo e 0,81 senza l'algoritmo) e di tumori più aggressivi (AUC = 0,87 con e 0,78 senza l'algoritmo). In uno studio condotto su 417 pazienti, una CNN ha ottenuto un’AUC di 0,81 per la classificazione del carcinoma prostatico clinicamente significativo utilizzando la RM multiparametrica, con una sensibilità solo leggermente inferiore a quella di un radiologo di grande esperienza (Cao et al., 2019).

Come per molte altre applicazioni dell’IA in radiologia, la mancanza di interpretabilità dei modelli di deep learning della RM della prostata ne ostacola e ritarda l’implementazione nella pratica clinica (Aristidou et al., 2022; Reddy et al., 2020; Reyes et al., 2020; Vayena et al., 2018). Uno studio che ha utilizzato una CNN sulla RM della prostata di 1.224 pazienti e le caratteristiche istopatologiche come riferimento ha rilevato un'AUC di 0,89 per la differenziazione del carcinoma prostatico clinicamente significativo da altre alterazioni della prostata (Hamm et al., 2023). Inoltre, gli autori hanno incluso una mappa termica voxel-wise delle aree sospette di carcinoma prostatico clinicamente significativo e spiegazioni descrittive, formulate sulla base della classificazione PIRADS, del modo in cui la CNN ha tratto la sua conclusione. L'algoritmo è stato associato a una riduzione del tempo di lettura da 85 secondi a 47 secondi e a un aumento della sicurezza nella lettura nei lettori non esperti.

Conclusione

L’imaging medico svolge un ruolo centrale negli iter di screening di molti dei tumori più comuni. La lettura degli esami di screening richiede competenza ed esperienza considerevoli e la domanda attuale supera di gran lunga l'offerta di radiologi formati (AAMC Report Reinforces Mounting Physician Shortage, 2021, Clinical Radiology UK Workforce Census 2019 Report, 2019). L’uso di strumenti basati sull’IA per lo screening oncologico promette di offrire un enorme contributo nella riduzione di queste problematiche. I vantaggi di tali approcci includono una migliore identificazione dei soggetti idonei allo screening, una migliore accuratezza diagnostica, una riduzione dei tempi di refertazione e una maggiore sicurezza da parte dei radiologi nelle decisioni diagnostiche. I risultati più promettenti sono stati ottenuti quando le decisioni sugli esami di screening sono state prese combinando i risultati dei sistemi basati sull’IA e l’interpretazione dei radiologi. Il processo decisionale fondato sulla collaborazione tra strumenti basati sull’IA e radiologi può quindi aprire la strada a una nuova era nello screening oncologico.

Bibliografia 

AAMC Report Reinforces Mounting Physician Shortage. (2021). AAMC. https://www.aamc.org/news-insights/press- releases/aamc-report-reinforces-mounting-physician-shortage

Aggarwal, R., Sounderajah, V., Martin, G., Ting, D. S. W., Karthikesalingam, A., King, D., Ashrafian, H., & Darzi, A. (2021). Diagnostic accuracy of deep learning in medical imaging: a systematic review and meta-analysis. NPJ Digital Medicine, 4(1), 65. https://doi.org/10.1038/s41746-021-00438-z

Ahmed, H. U., El-Shater Bosaily, A., Brown, L. C., Gabe, R., Kaplan, R., Parmar, M. K., Collaco-Moraes, Y., Ward, K., Hindley, R. G., Freeman, A., Kirkham, A. P., Oldroyd, R., Parker, C., Emberton, M., & PROMIS study group. (2017). Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study. The Lancet, 389(10071), 815–822. https://doi.org/10.1016/ S0140-6736(16)32401-1

AI for radiology. (n.d.). Retrieved June 4, 2023, from https://grand-challenge.org/aiforradiology/?subspeciality=All& modality=All&ce_under=All&ce_%20class=All&fda_ class=All&sort_by=ce+certification&search=prostate

Alkabbany, I., Ali, A. M., Mohamed, M., Elshazly, S. M., & Farag, A. (2022). An AI-Based Colonic Polyp Classifier for Colorectal Cancer Screening Using Low-Dose Abdominal CT. Sensors, 22, (24). https://doi.org/10.3390/s22249761

Al Mohammad, B., Hillis, S. L., Reed, W., Alakhras, M., & Brennan, P. C. (2019). Radiologist performance in the detection of lung cancer using CT. Clinical Radiology, 74(1), 67–75. https://doi.org/10.1016/j.crad.2018.10.008

Ardila, D., Kiraly, A. P., Bharadwaj, S., Choi, B., Reicher, J. J., Peng, L., Tse, D., Etemadi, M., Ye, W., Corrado, G., Naidich, D. P., & Shetty, S. (2019). End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nature Medicine, 25(6), 954–961. https://doi. org/10.1038/s41591-019-0447-x

Arif, M., Schoots, I. G., Castillo Tovar, J., Bangma, C. H., Krestin, G. P., Roobol, M. J., Niessen, W., & Veenland, J. F. (2020). Clinically significant prostate cancer detection and segmentation in low-risk patients using a convolutional neural network on multi-parametric MRI. European Radiology, 30(12), 6582–6592. https://doi.org/10.1007/s00330-020-07008-z

Aristidou, A., Jena, R., & Topol, E. J. (2022). Bridging the chasm between AI and clinical implementation. The Lancet, 399(10325), 620. https://doi.org/10.1016/S0140-6736(22)00235-5

Armato, S. G., 3rd, Roberts, R. Y., Kocherginsky, M., Aberle, D. R., Kazerooni, E. A., Macmahon, H., van Beek, E. J. R., Yankelevitz, D., McLennan, G., McNitt-Gray, M. F., Meyer, C. R., Reeves, A. P., Caligiuri, P., Quint, L. E., Sundaram, B., Croft, B. Y., & Clarke, L. P. (2009). Assessment of radiologist performance in the detection of lung nodules: dependence on the definition of "truth." Academic Radiology, 16(1), 28–38. https://doi.org/10.1016/j.acra.2008.05.022

Banks, E., Reeves, G., Beral, V., Bull, D., Crossley, B., Simmonds, M., Hilton, E., Bailey, S., Barrett, N., Briers, P., English, R., Jackson, A., Kutt, E., Lavelle, J., Rockall, L., Wallis, M. G., Wilson, M., & Patnick, J. (2006). Hormone replacement therapy and false positive recall in the Million Women Study: patterns of use, hormonal constituents and consistency of effect. Breast Cancer Research: BCR, 8(1), R8. https://doi.org/10.1186/bcr1364

Bardis, M., Houshyar, R., Chantaduly, C., Tran-Harding, K., Ushinsky, A., Chahine, C., Rupasinghe, M., Chow, D., & Chang, P. (2021). Segmentation of the Prostate Transition Zone and Peripheral Zone on MR Images with Deep Learning. Radiology. Imaging Cancer, 3(3), e200024. https://doi.org/10.1148/rycan.2021200024

Belue, M. J., & Turkbey, B. (2022). Tasks for artificial intelligence in prostate MRI. European Radiology Experimental, 6(1), 33. https://doi.org/10.1186/s41747-022-00287-9

Boyd, N. F., Guo, H., Martin, L. J., Sun, L., Stone, J., Fishell, E., Jong, R. A., Hislop, G., Chiarelli, A., Minkin, S., & Yaffe, M. J. (2007). Mammographic density and the risk and detection of breast cancer. The New England Journal of Medicine, 356(3), 227–236. https://doi.org/10.1056/NEJMoa062790

Bray, F., Ferlay, J., Soerjomataram, I., Siegel, R. L., Torre, L. A., & Jemal, A.(2018). Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: A Cancer Journal for Clinicians, 68(6), 394–424. https://doi.org/10.3322/caac.21492

Brehar, R., Mitrea, D.-A., Vancea, F., Marita, T., Nedevschi, S., Lupsor-Platon, M., Rotaru, M., & Badea, R. I. (2020). Comparison of Deep-Learning and Conventional Machine- Learning Methods for the Automatic Recognition of the Hepatocellular Carcinoma Areas from Ultrasound Images. Sensors, 20(11). https://doi.org/10.3390/s20113085

Cacciamani, G. E., Sanford, D. I., Chu, T. N., Kaneko, M., De Castro Abreu, A. L., Duddalwar, V., & Gill, I. S. (2023). Is Artificial Intelligence Replacing Our Radiology Stars? Not Yet! European Urology Open Science, 48, 14–16. https://doi. org/10.1016/j.euros.2022.09.024

Cao, R., Mohammadian Bajgiran, A., Afshari Mirak, S., Shakeri, S., Zhong, X., Enzmann, D., Raman, S., & Sung, K. (2019). Joint Prostate Cancer Detection and Gleason Score Prediction in mp-MRI via FocalNet. IEEE Transactions on Medical Imaging, 38 (11), 2496–2506. https://doi.org/10.1109/ TMI.2019.2901928

Castells, X., Molins, E., & Macià, F. (2006). Cumulative false positive recall rate and association with participant related factors in a population based breast cancer screening programme. Journal of Epidemiology and Community Health, 60(4), 316–321. https://doi.org/10.1136/jech.2005.042119

Chamberlin, J., Kocher, M. R., Waltz, J., Snoddy, M., Stringer, N. F. C., Stephenson, J., Sahbaee, P., Sharma, P., Rapaka, S., Schoepf, U. J., Abadia, A. F., Sperl, J., Hoelzer, P., Mercer, M., Somayaji, N., Aquino, G., & Burt, J. R. (2021). Automated detection of lung nodules and coronary artery calcium using artificial intelligence on low-dose CT scans for lung cancer screening: accuracy and prognostic value. BMC Medicine, 19(1), 55. https://doi.org/10.1186/s12916-021-01928-3

Clinical radiology UK workforce census 2019 report. (2019). https://www.rcr.ac.uk/publication/clinical-radiology-uk- workforce-census-2019-report

Colli, A., Fraquelli, M., Casazza, G., Massironi, S., Colucci, A., Conte, D., & Duca, P. (2006). Accuracy of ultrasonography, spiral CT, magnetic resonance, and alpha-fetoprotein in diagnosing hepatocellular carcinoma: a systematic review. The American Journal of Gastroenterology, 101(3), 513–523. https://doi.org/10.1111/j.1572-0241.2006.00467

Dembrower, K., Wåhlin, E., Liu, Y., Salim, M., Smith, K., Lindholm, P., Eklund, M., & Strand, F. (2020). Effect of artificial intelligence-based triaging of breast cancer screening mammograms on cancer detection and radiologist workload: a retrospective simulation study. The Lancet. Digital Health, 2(9), e468–e474. https://doi.org/10.1016/S2589-7500(20)30185-0

Drost, F.-J. H., Osses, D. F., Nieboer, D., Steyerberg, E. W., Bangma, C. H., Roobol, M. J., & Schoots, I. G.(2019). Prostate MRI, with or without MRI-targeted biopsy, and systematic biopsy for detecting prostate cancer. Cochrane Database of Systematic Reviews, 4(4), CD012663. https://doi.org/10.1002/14651858. CD012663.pub2

Elmore, J. G., Jackson, S. L., Abraham, L., Miglioretti, D. L., Carney, P. A., Geller, B. M., Yankaskas, B. C., Kerlikowske, K., Onega, T., Rosenberg, R. D., Sickles, E. A., & Buist, D. S. M. (2009). Variability in interpretive performance at screening mammography and radiologists’ characteristics associated with accuracy. Radiology, 253(3), 641–651. https://doi.org/10.1148/radiol.2533082308

Elwenspoek, M. M. C., Sheppard, A. L., McInnes, M. D. F., Merriel, S. W. D., Rowe, E. W. J., Bryant, R. J., Donovan, J. L., & Whiting, P. (2019). Comparison of Multiparametric Magnetic Resonance Imaging and Targeted Biopsy With Systematic Biopsy Alone for the Diagnosis of Prostate Cancer: A Systematic Review and Meta-analysis. JAMA Network Open, 2(8), e198427. https:// doi.org/10.1001/jamanetworkopen.2019.8427

European Association for the Study of the Liver. (2018). EASL Clinical Practice Guidelines: Management of hepatocellular carcinoma. Journal of Hepatology, 69(1), 182–236. https://doi.org/10.1016/j.jhep.2018.03.019

Existence of national screening program for breast cancer. (n.d.). Retrieved April 2, 2023, from https://www.who.int/data/gho/data/indicators/indicator-details/GHO/existence-ofnational-screening-program-for-breast-cancer

Expert Panel on Gastrointestinal Imaging:, Moreno, C., Kim, D. H., Bartel, T. B., Cash, B. D., Chang, K. J., Feig, B. W., Fowler, K. J., Garcia, E. M., Kambadakone, A. R., Lambert, D. L., Levy, A. D., Marin, D., Peterson, C. M., Scheirey, C. D., Smith, M. P., Weinstein, S., & Carucci, L. R. (2018). ACR Appropriateness Criteria® Colorectal Cancer Screening. Journal of the American College of Radiology: JACR, 15(5S), S56–S68. https://doi.org/10.1016/j.jacr.2018.03.014

Ferlay, J., Colombet, M., Soerjomataram, I., Dyba, T., Randi,G., Bettio, M., Gavin, A., Visser, O., & Bray, F.(2018). Cancer incidence and mortality patterns in Europe: Estimates for 40 countries and 25 major cancers in 2018. European Journal of Cancer, 103, 356–387. https://doi.org/10.1016/j.ejca.2018.07.005

Freeman, K., Geppert, J., Stinton, C., Todkill, D., Johnson, S., Clarke, A., & Taylor-Phillips, S. (2021). Use of artificial intelligence for image analysis in breast cancer screening programmes: systematic review of test accuracy.BMJ, 374, n1872.https://doi.org/10.1136/bmj.n1872

Frenette, C. T., Isaacson, A. J., Bargellini, I., Saab, S., & Singal, A. G. (2019). A Practical Guideline for Hepatocellular Carcinoma Screening in Patients at Risk. Mayo Clinic Proceedings. Innovations, Quality & Outcomes, 3(3), 302–310. https://doi.org/10.1016/j.mayocpiqo.2019.04.005

Garvey, B., Türkbey, B., Truong, H., Bernardo, M., Periaswamy, S., & Choyke, P. L. (2014). Clinical value of prostate segmentation and volume determination on MRI in benign prostatic hyperplasia. Diagnostic and Interventional Radiology, 20(3), 229–233. https://doi.org/10.5152/dir.2014.13322

Gatos, I., Tsantis, S., Spiliopoulos, S., Skouroliakou, A., Theotokas, I., Zoumpoulis, P., Hazle, J. D., & Kagadis, G. C. (2015). A new automated quantification algorithm for the detection and evaluation of focal liver lesions with contrastenhanced ultrasound. Medical Physics, 42(7), 3948–3959. https://doi.org/10.1118/1.4921753

Gaur, S., Lay, N., Harmon, S. A., Doddakashi, S., Mehralivand, S., Argun, B., Barrett, T., Bednarova, S., Girometti, R., Karaarslan, E., Kural, A. R., Oto, A., Purysko, A. S., Antic, T., Magi-Galluzzi, C., Saglican, Y., Sioletic, S., Warren, A. Y., Bittencourt, L., … Turkbey, B. (2018). Can computeraided diagnosis assist in the identification of prostate cancer on prostate MRI? a multi-center, multi-reader investigation. Oncotarget, 9(73), 33804–33817. https://doi.org/10.18632/ oncotarget.26100

Giannini, V., Mazzetti, S., Defeudis, A., Stranieri, G., Calandri, M., Bollito, E., Bosco, M., Porpiglia, F., Manfredi, M., De Pascale, A., Veltri, A., Russo, F., & Regge, D. (2021). A Fully Automatic Artificial Intelligence System Able to Detect and Characterize Prostate Cancer Using Multiparametric MRI: Multicenter and Multi-Scanner Validation. Frontiers in Oncology, 11, 718155. https://doi.org/10.3389/fonc.2021.718155

Gierada, D. S., Pinsky, P. F., Duan, F., Garg, K., Hart, E. M., Kazerooni, E. A., Nath, H., Watts, J. R., Jr, & Aberle, D. R. (2017). Interval lung cancer after a negative CT screening examination: CT findings and outcomes in National Lung Screening Trial participants. European Radiology, 27(8), 3249–3256. https://doi.org/10.1007/s00330-016-4705-8

Giordano, L., von Karsa, L., Tomatis, M., Majek, O., de Wolf, C., Lancucki, L., Hofvind, S., Nyström, L., Segnan, N., Ponti, A., Eunice Working Group, Van Hal, G., Martens, P., Májek, O., Danes, J., von Euler-Chelpin, M., Aasmaa, A., Anttila, A., Becker, N., … Suonio, E. (2012). Mammographic screening programmes in Europe: organization, coverage and participation. Journal of Medical Screening, 19 Suppl 1, 72–82. https://doi.org/10.1258/jms.2012.012085

Grosu, S., Wesp, P., Graser, A., Maurus, S., Schulz, C., Knösel, T., Cyran, C. C., Ricke, J., Ingrisch, M., & Kazmierczak, P. M. (2021). Machine Learning-based Differentiation of Benign and Premalignant Colorectal Polyps Detected with CT Colonography in an Asymptomatic Screening Population: A Proof-of-Concept Study. Radiology, 299(2), 326–335. https://doi.org/10.1148/radiol.2021202363

Hamm, C. A., Baumgärtner, G. L., Biessmann, F., Beetz, N. L., Hartenstein, A., Savic, L. J., Froböse, K., Dräger, F., Schallenberg, S., Rudolph, M., Baur, A. D. J., Hamm, B., Haas, M., Hofbauer, S., Cash, H., & Penzkofer, T.(2023). Interactive Explainable Deep Learning Model Informs Prostate Cancer Diagnosis at MRI. Radiology, 307(4), e222276. https://doi.org/10.1148/radiol.222276

Hassan, T. M., Elmogy, M., & Sallam, E.-S. (2017). Diagnosis of Focal Liver Diseases Based on Deep Learning Technique for Ultrasound Images. Arabian Journal for Science and Engineering, 42(8), 3127–3140. https://doi.org/10.1007/s13369-016-2387-9

Heiken, J. P. (2007). Distinguishing benign from malignant liver tumours. Cancer Imaging: The Official Publication of the International Cancer Imaging Society, 7 Spec No A(Special issue A), S1–S14. https://doi.org/10.1102/1470-7330.2007.9084

Helsingen Lise M., & Kalager Mette. (2022). Colorectal Cancer Screening — Approach, Evidence, and Future Directions. NEJM Evidence, 1(1), EVIDra2100035. https://doi.org/10.1056/ EVIDra2100035

Hoeks, C. M. A., Schouten, M. G., Bomers, J. G. R., Hoogendoorn, S. P., Hulsbergen-van de Kaa, C. A., Hambrock, T., Vergunst, H., Sedelaar, J. P. M., Fütterer, J. J., & Barentsz, J. O. (2012). Three-Tesla magnetic resonance-guided prostate biopsy in men with increased prostate-specific antigen and repeated, negative, random, systematic, transrectal ultrasound biopsies: detection of clinically significant prostate cancers. European Urology, 62(5), 902–909. https://doi.org/10.1016/j. eururo.2012.01.047

Huang, Q., Pan, F., Li, W., Yuan, F., Hu, H., Huang, J., Yu, J., & Wang, W. (2020). Differential Diagnosis of Atypical Hepatocellular Carcinoma in Contrast-Enhanced Ultrasound Using Spatio-Temporal Diagnostic Semantics. IEEE Journal of Biomedical and Health Informatics, 24(10), 2860–2869. https://doi.org/10.1109/JBHI.2020.2977937

Hugosson, J., Månsson, M., Wallström, J., Axcrona, U., Carlsson, S. V., Egevad, L., Geterud, K., Khatami, A., Kohestani, K., Pihl, C.-G., Socratous, A., Stranne, J., Godtman, R. A., Hellström, M., & GÖTEBORG-2 Trial Investigators. (2022). Prostate Cancer Screening with PSA and MRI Followed by Targeted Biopsy Only.The New England Journal of Medicine, 387(23), 2126–2137. https://doi.org/10.1056/NEJMoa2209454

Inadomi, J. M., Vijan, S., Janz, N. K., Fagerlin, A., Thomas, J. P., Lin, Y. V., Muñoz, R., Lau, C., Somsouk, M., El-Nachef, N., & Hayward, R. A. (2012). Adherence to colorectal cancer screening: a randomized clinical trial of competing strategies. Archives of Internal Medicine, 172(7), 575–582. https://doi. org/10.1001/archinternmed. 2012.332

Joseph, D. A., Jessica B. King, M. P. H., Miller, J. W., & Richardson, L. C. (2012, June 15).Prevalence of Colorectal Cancer Screening Among Adults — Behavioral Risk Factor Surveillance System, United States, 2010. https://www.cdc.gov/MMWr/preview/mmwrhtml/su6102a9.htm

Katase, S., Ichinose, A., Hayashi, M., Watanabe, M., Chin,K., Takeshita, Y., Shiga, H., Tateishi, H., Onozawa, S., Shirakawa, Y., Yamashita, K., Shudo, J., Nakamura, K., Nakanishi, A., Kuroki, K., & Yokoyama, K. (2022). Development and performance evaluation of a deep learning lung nodule detection system. BMC Medical Imaging. 22(1), 203. https://doi.org/10.1186/s12880-022-00938-8

Kinsinger, L. S., Anderson, C., Kim, J., Larson, M., Chan, S. H., King, H. A., Rice, K. L., Slatore, C. G., Tanner, N. T., Pittman, K., Monte, R. J., McNeil, R. B., Grubber, J. M., Kelley, M. J., Provenzale, D., Datta, S. K., Sperber, N. S., Barnes, L. K., Abbott, D. H., … Jackson, G. L. (2017). Implementation of Lung Cancer Screening in the Veterans Health Administration. JAMA Internal Medicine, 177(3), 399–406. https://doi.org/10.1001/jamainternmed.2016.9022

Klotz, L., Vesprini, D., Sethukavalan, P., Jethava, V., Zhang, L., Jain, S., Yamamoto, T., Mamedov, A., & Loblaw, A. (2015). Long-term follow-up of a large active surveillance cohort of patients with prostate cancer.Journal of Clinical Oncology: Official Journal of the American Society of Clinical Oncology, 33(3), 272–277.https://doi.org/10.1200/JCO.2014.55.1192

Kohestani, K., Wallström, J., Dehlfors, N., Sponga, O. M., Månsson, M., Josefsson, A., Carlsson, S., Hellström, M., & Hugosson, J. (2019). Performance and inter-observer variability of prostate MRI (PI-RADS version 2) outside high-volume centres. Scandinavian Journal of Urology, 53(5), 304–311. https://doi.org/10.1080/21681805.2019.1675757

Kondo, S., Takagi, K., Nishida, M., Iwai, T., Kudo, Y., Ogawa, K., Kamiyama, T., Shibuya, H., Kahata, K., & Shimizu, C. (2017). Computer-Aided Diagnosis of Focal Liver Lesions Using Contrast-Enhanced Ultrasonography With Perflubutane Microbubbles. IEEE Transactions on Medical Imaging, 36(7),1427–1437. https://doi.org/10.1109/TMI.2017.2659734

Kramer, B. S. (2004). The science of early detection. Urologic Oncology, 22(4), 344–347. https://doi.org/10.1016/j.urolonc.2003.04.001

Kudo, M. (2009). Multistep human hepatocarcinogenesis: correlation of imaging with pathology. Journal of Gastroenterology, 44 Suppl 19, 112–118. https://doi.org/10.1007/s00535-008-2274-6

Kuipers, E. J., Grady, W. M., Lieberman, D., Seufferlein, T., Sung, J. J., Boelens, P. G., van de Velde, C. J. H., & Watanabe, T. (2015) Colorectal cancer.Nature Reviews. Disease Primers, 1, 15065. https://doi.org/10.1038/nrdp.2015.65

Lam, D. L., Pandharipande, P. V., Lee, J. M., Lehman, C. D., & Lee, C. I. (2014). Imaging-based screening: understanding the controversies.AJR. American Journal of Roentgenology, 203(5), 952–956. https://doi.org/10.2214/AJR.14.13049

Lauritzen, A. D., Rodríguez-Ruiz, A., von Euler-Chelpin, M. C., Lynge, E., Vejborg, I., Nielsen, M., Karssemeijer, N., & Lillholm, M. (2022). An Artificial Intelligence-based Mammography Screening Protocol for Breast Cancer: Outcome and Radiologist Workload. Radiology, 304(1), 41–49. https://doi.org/10.1148/radiol.210948

Leader, J. K., Warfel, T. E., Fuhrman, C. R., Golla, S. K., Weissfeld, J. L., Avila, R. S., Turner, W. D., & Zheng, B. (2005). Pulmonary nodule detection with low-dose CT of the lung: agreement among radiologists. AJR. American Journal of Roentgenology, 185(4), 973–978. https://doi.org/10.2214/AJR.04.1225

Lehman, C. D., Wellman, R. D., Buist, D. S. M., Kerlikowske, K., Tosteson, A. N. A., Miglioretti, D. L., & Breast Cancer Surveillance Consortium. (2015). Diagnostic Accuracy of Digital Screening Mammography With and Without Computer-Aided Detection. JAMA Internal Medicine, 175(11), 1828–1837. https://doi.org/10.1001/jamainternmed.2015.5231

Leibig, C., Brehmer, M., Bunk, S., Byng, D., Pinker, K., & Umutlu, L. (2022). Combining the strengths of radiologists and AI for breast cancer screening: a retrospective analysis. The Lancet. Digital Health, 4(7), e507–e519. https://doi.org/10.1016/ S2589-7500(22)00070-X

Leitlinienprogramm Onkologie: Prostatakarzinom. (n.d.). Retrieved July 8, 2023, from https://www.leitlinienprogrammonkologie. de/index.php?id=58&type=0

Litjens, G. J. S., Barentsz, J. O., Karssemeijer, N., & Huisman, H. J. (2015). Clinical evaluation of a computer-aided diagnosis system for determining cancer aggressiveness in prostate MRI. European Radiology, 25(11), 3187–3199. https://doi.org/10.1007/s00330-015-3743-y

Lu, M. T., Raghu, V. K., Mayrhofer, T., Aerts, H. J. W. L., & Hoffmann, U. (2020). Deep Learning Using Chest Radiographs to Identify High-Risk Smokers for Lung Cancer Screening Computed Tomography: Development and Validation of a Prediction Model. Annals of Internal Medicine, 173(9), 704–713. https://doi.org/10.7326/M20-1868

Lung cancer: Screening. (2021, March 9). US Preventive Services Taskforce. https://www.uspreventiveservicestaskforce.org/uspstf/recommendation/lung-cancer-screening

Marmot, M. G., Altman, D. G., Cameron, D. A., Dewar, J. A., Thompson, S. G., & Wilcox, M. (2013). The benefits and harms of breast cancer screening: an independent review. British Journal of Cancer, 108(11), 2205–2240. https://doi.org/10.1038/bjc.2013.177

Marrero, J. A., Kulik, L. M., Sirlin, C. B., Zhu, A. X., Finn, R. S., Abecassis, M. M., Roberts, L. R., & Heimbach, J. K. (2018). Diagnosis, Staging, and Management of Hepatocellular Carcinoma: 2018 Practice Guidance by the American Association for the Study of Liver Diseases. Hepatology, 68(2), 723–750. https://doi.org/10.1002/hep.29913

McKinney, S. M., Sieniek, M., Godbole, V., Godwin, J., Antropova, N., Ashrafian, H., Back, T., Chesus, M., Corrado, G. S., Darzi, A., Etemadi, M., Garcia-Vicente, F., Gilbert, F. J., Halling-Brown, M., Hassabis, D., Jansen, S., Karthikesalingam, A., Kelly, C. J., King, D., … Shetty, S. (2020). International evaluation of an AI system for breast cancer screening. Nature, 577(7788), 89–94. https://doi.org/10.1038/s41586-019-1799-6

Merriel, S. W. D., Pocock, L., Gilbert, E., Creavin, S., Walter, F. M., Spencer, A., & Hamilton, W. (2022). Systematic review and meta-analysis of the diagnostic accuracy of prostatespecific antigen (PSA) for the detection of prostate cancer in symptomatic patients. BMC Medicine, 20(1), 54. https://doi.org/10.1186/s12916-021-02230-y

Mikhael, P. G., Wohlwend, J., Yala, A., Karstens, L., Xiang, J., Takigami, A. K., Bourgouin, P. P., Chan, P., Mrah, S., Amayri, W., Juan, Y.-H., Yang, C.-T., Wan, Y.-L., Lin, G., Sequist, L. V., Fintelmann, F. J., & Barzilay, R. (2023). Sybil: A Validated Deep Learning Model to Predict Future Lung Cancer Risk From a Single Low-Dose Chest Computed Tomography.Journal of Clinical Oncology: Official Journal of the American Society of Clinical Oncology, 41(12), 2191–2200. https://doi.org/10.1200/JCO.22.01345

Moawad, F. J., Maydonovitch, C. L., Cullen, P. A., Barlow, D.S., Jenson, D. W., & Cash, B. D. 2010). CT colonography may improve colorectal cancer screening compliance. AJR. American Journal of Roentgenology, 195(5), 1118–1123. https://doi.org/10.2214/AJR.10.4921

Moran, S., & Warren-Forward, H. (2012). The Australian BreastScreen workforce: a snapshot. The Radiographer, 59(1), 26–30. https://doi.org/10.1002/j.2051-3909.2012.tb00169.x

Mottet, N., Bellmunt, J., Bolla, M., Briers, E., Cumberbatch, M. G., De Santis, M., Fossati, N., Gross, T., Henry, A. M., Joniau, S., Lam, T. B., Mason, M. D., Matveev, V. B., Moldovan, P. C., van den Bergh, R. C. N., Van den Broeck, T., van der Poel, H. G., van der Kwast, T. H., Rouvière, O., … Cornford, P. (2017). EAU-ESTRO-SIOG Guidelines on Prostate Cancer. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. European Urology 71(4), 618–629. https://doi.org/10.1016/j.eururo.2016.08.003

Muller, B. G., Shih, J. H., Sankineni, S., Marko, J., Rais-Bahrami, S., George, A. K., de la Rosette, J. J. M. C. H., Merino, M. J., Wood, B. J., Pinto, P., Choyke, P. L., & Turkbey, B. (2015). Prostate Cancer: Interobserver Agreement and Accuracy with the Revised Prostate Imaging Reporting and Data System at Multiparametric MR Imaging. Radiology, 277(3), 741–750. https://doi.org/10.1148/radiol.2015142818

National Lung Screening Trial Research Team, Aberle, D. R., Adams, A. M., Berg, C. D., Black, W. C., Clapp, J. D., Fagerstrom, R. M., Gareen, I. F., Gatsonis, C., Marcus, P.M., & Sicks, J. D. (2011). Reduced lung-cancer mortality with low-dose computed tomographic screening. The New England Journal of Medicine, 365(5), 395–409. https://doi.org/10.1056/NEJMoa1102873

NHS England. (2022). Targeted screening for lung cancer with low radiation dose computed tomography Standard protocol prepared for the Targeted Lung Health Checks Programme. https://www.england.nhs.uk/publication/targeted-screening-for-lung-cancer

OECD. (2012). Screening, survival and mortality for colorectal cancer.In Health at a Glance: Europe 2012 (pp. 110–111). OECD. https://doi.org/10.1787/9789264183896-48-en

Overview | Prostate cancer: diagnosis and management | Guidance | NICE. (n.d.). Retrieved July 8, 2023, from https://www.nice.org.uk/guidance/ng131

Park, H. J., Choi, B. I., Lee, E. S., Park, S. B., & Lee, J. B. (2017). How to Differentiate Borderline Hepatic Nodules in Hepatocarcinogenesis: Emphasis on Imaging Diagnosis. Liver Cancer, 6(3), 189–203. https://doi.org/10.1159/000455949

Penzkofer, T., Tuncali, K., Fedorov, A., Song, S.-E., Tokuda, J., Fennessy, F. M., Vangel, M. G., Kibel, A. S., Mulkern, R. V., Wells, W. M., Hata, N., & Tempany, C. M. C. (2015). Transperineal in-bore 3-T MR imaging-guided prostate biopsy: a prospective clinical observational study.Radiology, 274(1), 170–180. https://doi.org/10.1148/radiol.14140221

Pickhardt, P. J., Choi, J. R., Hwang, I., Butler, J. A., Puckett, M.L., Hildebrandt, H. A., Wong, R. K., Nugent, P. A., Mysliwiec, P. A., & Schindler, W. R. (2003). Computed tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults.The New England Journal of Medicine, 349(23), 2191–2200. https://doi.org/10.1056/NEJMoa031618

Pickhardt, P. J., Correale, L., Delsanto, S., Regge, D., & Hassan, C. (2018). CT Colonography Performance for the Detection of Polyps and Cancer in Adults ≥ 65 Years Old: Systematic Review and Meta-Analysis.AJR. American Journal of Roentgenology, 211(1), 40–51. https://doi.org/10.2214/AJR.18.19515

Pickhardt, P. J., Hassan, C., Halligan, S., & Marmo, R. (2011). Colorectal cancer: CT colonography and colonoscopy for detection-systematic review and meta-analysis.Radiology, 259(2), 393–405. https://doi.org/10.1148/radiol.11101887

Raya-Povedano, J. L., Romero-Martín, S., Elías-Cabot, E., Gubern-Mérida, A., Rodríguez-Ruiz, A., & Álvarez-Benito, M. (2021). AI-based Strategies to Reduce Workload in Breast Cancer Screening with Mammography and Tomosynthesis: A Retrospective Evaluation. Radiology, 300(1), 57–65. https://doi. org/10.1148/radiol.2021203555

Reddy, S., Allan, S., Coghlan, S., & Cooper, P. (2020). A governance model for the application of AI in health care. Journal of the American Medical Informatics Association: JAMIA, 27(3), 491–497. https://doi.org/10.1093/jamia/ocz192

Ren, W., Chen, M., Qiao, Y., & Zhao, F. (2022). Global guidelines for breast cancer screening: A systematic review.Breast, 64,85–99. https://doi.org/10.1016/j.breast.2022.04.003

Reyes, M., Meier, R., Pereira, S., Silva, C. A., Dahlweid, F.-M., von Tengg-Kobligk, H., Summers, R. M., & Wiest, R. (2020). On the Interpretability of Artificial Intelligence in Radiology: Challenges and Opportunities. Radiology.Artificial Intelligence, 2(3), e190043. https://doi.org/10.1148/ryai.2020190043

Rimmer, A. (2017). Radiologist shortage leaves patient care at risk, warns royal college.BMJ, 359, j4683. https://doi.org/10.1136/bmj.j4683

Ristvedt, S. L., McFarland, E. G., Weinstock, L. B., & Thyssen, E. P. (2003). Patient preferences for CT colonography, conventional colonoscopy, and bowel preparation. The American Journal of Gastroenterology, 98(3), 578–585.https://doi.org/10.1016/S0002-9270(02)06024-0

Rosenkrantz, A. B., Ayoola, A., Hoffman, D., Khasgiwala, A., Prabhu, V., Smereka, P., Somberg, M., & Taneja, S. S. (2017). The Learning Curve in Prostate MRI Interpretation: Self-Directed Learning Versus Continual Reader Feedback. AJR. American Journal of Roentgenology, 208(3), W92–W100. https://doi.org/10.2214/AJR.16.16876

Saha, A., Hosseinzadeh, M., & Huisman, H. (2021). End-to-end prostate cancer detection in bpMRI via 3D CNNs: Effects of attention mechanisms, clinical priori and decoupled false positive reduction. Medical Image Analysis, 73, 102155. https://doi.org/10.1016/j.media.2021.102155

Sanford, T. H., Zhang, L., Harmon, S. A., Sackett, J., Yang, D., Roth, H., Xu, Z., Kesani, D., Mehralivand, S., Baroni, R.H., Barrett, T., Girometti, R., Oto, A., Purysko, A. S., Xu, S., Pinto, P. A., Xu, D., Wood, B. J., Choyke, P. L., & Turkbey, B. (2020). Data Augmentation and Transfer Learning to Improve Generalizability of an Automated Prostate Segmentation Model. AJR. American Journal of Roentgenology, 215(6), 1403–1410. https://doi.org/10.2214/AJR.19.22347

Sato, M., Kobayashi, T., Soroida, Y., Tanaka, T., Nakatsuka,T., Nakagawa, H., Nakamura, A., Kurihara, M., Endo, M., Hikita, H., Sato, M., Gotoh, H., Iwai, T., Tateishi, R., Koike,K., & Yatomi, Y. (2022). Development of novel deep multimodal representation learning-based model for the differentiation of liver tumors on B-mode ultrasound images. Journal of Gastroenterology and Hepatology, 37(4), 678–684, https://doi.org/10.1111/jgh.15763

Schmauch, B., Herent, P., Jehanno, P., Dehaene, O., Saillard, C., Aubé, C., Luciani, A., Lassau, N., & Jégou, S. (2019). Diagnosis of focal liver lesions from ultrasound using deep learning.Diagnostic and Interventional Imaging, 100(4), 227–233. https://doi.org/10.1016/j.diii.2019.02.009

Siddiqui, M. M., Rais-Bahrami, S., Turkbey, B., George, A. K., Rothwax, J., Shakir, N., Okoro, C., Raskolnikov, D., Parnes, H.L., Linehan, W. M., Merino, M. J., Simon, R. M., Choyke, P. L., Wood, B. J., & Pinto, P. A. (2015). Comparison of MR/ultrasound fusion-guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA: The Journal of the American Medical Association, 313(4), 390–397. https://doi.org/10.1001/jama.2014.17942

Siegel, R. L., Miller, K. D., Fuchs, H. E., & Jemal, A. (2021). Cancer Statistics, 2021. CA: A Cancer Journal for Clinicians, 71(1), 7–33. https://doi.org/10.3322/caac.21654

Singal, A. G., Zhang, E., Narasimman, M., Rich, N. E., Waljee, A. K., Hoshida, Y., Yang, J. D., Reig, M., Cabibbo, G., Nahon, P., Parikh, N. D., & Marrero, J.A. (2022). HCC surveillance improves early detection, curative treatment receipt, and survival in patients with cirrhosis: A meta-analysis Journal of Hepatology, 77(1), 128–139. https://doi.org/10.1016/j.jhep.2022.01.023

Smith, C. P., Harmon, S. A., Barrett, T., Bittencourt, L. K., Law, Y. M., Shebel, H., An, J. Y., Czarniecki, M., Mehralivand, S.,Coskun, M., Wood, B. J., Pinto, P. A., Shih, J. H., Choyke, P. L., & Turkbey, B. (2019). Intra- and interreader reproducibility of PI-RADSv2: A multireader study. Journal of Magnetic Resonance Imaging: JMRI, 49(6), 1694–1703. https://doi.org/10.1002/jmri.26555

Smyth, A. E., Healy, C. F., MacMathuna, P., & Fenlon, H. M. (2013). REVIEW OF CT COLONOGRAPHY: REAL-LIFE EXPERIENCE OF ONE THOUSAND CASES IN A TERTIARY REFERRAL CENTRE. Gut, 62(Suppl 2), A15–A15. https://doi.org/10.1136/gutjnl-2013-305143.36

Song, B., Zhang, G., Lu, H., Wang, H., Zhu, W., J Pickhardt, P., & Liang, Z. (2014). Volumetric texture features from higher-order images for diagnosis of colon lesions via CT colonography. International Journal of Computer Assisted Radiology and Surgery, 9(6), 1021–1031. https://doi.org/10.1007/s11548-014-0991-2

Sonn, G. A., Chang, E., Natarajan, S., Margolis, D. J., Macairan, M., Lieu, P., Huang, J., Dorey, F. J., Reiter, R.E., & Marks, L. S. (2014). Value of targeted prostate biopsy using magnetic resonance-ultrasound fusion in men with prior negative biopsy and elevated prostate-specific antigen. European Urology, 65(4), 809–815. https://doi.org/10.1016/j. eururo.2013.03.025

Stock, C., Ihle, P., Schubert, I., & Brenner, H. (2011). Colonoscopy and fecal occult blood test use in Germany: results from a large insurance-based cohort. Endoscopy, 43(9), 771–781. https://doi.org/10.1055/s-0030-1256504

Sung, H., Ferlay, J., Siegel, R. L., Laversanne, M., Soerjomataram, I., Jemal, A., & Bray, F. (2021). Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA: A Cancer Journal for Clinicians, 71(3), 209–249. https://doi.org/10.3322/caac.21660

Sunoqrot, M. R. S., Saha, A., Hosseinzadeh, M., Elschot, M., & Huisman, H. (2022). Artificial intelligence for prostate MRI: open datasets, available applications, and grand challenges.European Radiology Experimental, 6(1), 35. https://doi.org/10.1186/s41747-022-00288-8

Tabár, L., Vitak, B., Chen, T. H.-H., Yen, A. M.-F., Cohen, A., Tot, T., Chiu, S. Y.-H., Chen, S. L.-S., Fann, J. C.-Y., Rosell, J., Fohlin, H., Smith, R. A., & Duffy, S. W. (2011). Swedish two-county trial: impact of mammographic screening on breast cancer mortality during 3 decades. Radiology, 260(3), 658–663. https://doi.org/10.1148/radiol.11110469

Ta, C. N., Kono, Y., Eghtedari, M., Oh, Y. T., Robbin, M. L., Barr, R. G., Kummel, A. C., & Mattrey, R. F. (2018). Focal Liver Lesions: Computer-aided Diagnosis by Using Contrast-enhanced US Cine Recordings. Radiology, 286(3), 1062–1071. https://doi.org/10.1148/radiol.2017170365

The burden of cancer. (n.d.). The Cancer Atlas. Retrieved July 8, 2023, from https://canceratlas.cancer.org/the-burden/theburden-of-cancer/

Tiyarattanachai, T., Apiparakoon, T., Marukatat, S., Sukcharoen, S., Yimsawad, S., Chaichuen, O., Bhumiwat, S., Tanpowpong, N., Pinjaroen, N., Rerknimitr, R., & Chaiteerakij, R. (2022). The feasibility to use artificial intelligence to aid detecting focal liver lesions in real-time ultrasound: a preliminary study based on videos. Scientific Reports 12(1), 7749. https://doi.org/10.1038/s41598-022-11506-z

Turco, S., Tiyarattanachai, T., Ebrahimkheil, K., Eisenbrey, J., Kamaya, A., Mischi, M., Lyshchik, A., & Kaffas, A. E. (2022). Interpretable Machine Learning for Characterization of Focal Liver Lesions by Contrast-Enhanced Ultrasound. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 69(5), 1670–1681. https://doi.org/10.1109/TUFFC.2022.3161719

Turkbey, B., & Haider, M. A. (2022). Deep learning-based artificial intelligence applications in prostate MRI: brief summary. The British Journal of Radiology, 95(1131), 20210563. https://doi.org/10.1259/bjr.20210563

Twilt, J. J., van Leeuwen, K. G., Huisman, H. J., Fütterer, J. J., & de Rooij, M. (2021). Artificial Intelligence Based Algorithms for Prostate Cancer Classification and Detection on Magnetic Resonance Imaging: A Narrative Review.Diagnostics (Basel, Switzerland), 11(6). https://doi.org/10.3390/diagnostics11060959

Tzartzeva, K., Obi, J., Rich, N. E., Parikh, N. D., Marrero, J.A., Yopp, A., Waljee, A. K., & Singal, A. G. (2018). Surveillance Imaging and Alpha Fetoprotein for Early Detection of Hepatocellular Carcinoma in Patients With Cirrhosis: A Meta-analysis. Gastroenterology, 154(6), 1706–1718.e1. https://doi.org/10.1053/j.gastro.2018.01.064

Use of colorectal cancer screening tests. (2023, March 31). https://www.cdc.gov/cancer/colorectal/statistics/use-screeningtests-BRFSS.htm

Ushinsky, A., Bardis, M., Glavis-Bloom, J., Uchio, E., Chantaduly, C., Nguyentat, M., Chow, D., Chang, P. D., & Houshyar, R. (2021). A 3D-2D Hybrid U-Net Convolutional Neural Network Approach to Prostate Organ Segmentation of Multiparametric MRI.AJR. American Journal of Roentgenology, 216(1), 111–116. https://doi.org/10.2214/AJR.19.22168

US Preventive Services Task Force, Grossman, D. C., Curry, S. J., Owens, D. K., Bibbins-Domingo, K., Caughey, A. B., Davidson, K. W., Doubeni, C. A., Ebell, M., Epling, J. W., Jr, Kemper, A. R., Krist, A. H., Kubik, M., Landefeld, C. S., Mangione, C. M., Silverstein, M., Simon, M. A., Siu, A. L., & Tseng, C.-W. (2018). Screening for Prostate Cancer: US Preventive Services Task Force Recommendation Statement. JAMA: The Journal of the American Medical Association, 319(18), 1901–1913. https://doi.org/10.1001/jama.2018.3710

van Leeuwen, K. G., Schalekamp, S., Rutten, M. J. C. M., van Ginneken, B., & de Rooij, M. (2021). Artificial intelligence in radiology: 100 commercially available products and their scientific evidence. European Radiology, 31(6), 3797–3804. https://doi.org/10.1007/s00330-021-07892-z

Vayena, E., Blasimme, A., & Cohen, I. G. (2018). Machine learning in medicine: Addressing ethical challenges.PLoS Medicine, 15(11), e1002689. https://doi.org/10.1371/journal.pmed.1002689

Virmani, J., Kumar, V., Kalra, N., & Khandelwal, N. (2014).Neural network ensemble based CAD system for focal liver lesions from B-mode ultrasound. Journal of Digital Imaging, 27(4), 520–537. https://doi.org/10.1007/s10278-014-9685-0

Vogel, A., Meyer, T., Sapisochin, G., Salem, R., & Saborowski, A. (2022). Hepatocellular carcinoma.The Lancet, 400(10360), 1345–1362. https://doi.org/10.1016/S0140-6736(22)01200-4

Walker, S. M., Choyke, P. L., & Turkbey, B. (2020). What You Need to Know Before Reading Multiparametric MRI for Prostate Cancer. AJR. American Journal of Roentgenology, 214(6), 1211–1219. https://doi.org/10.2214/AJR.19.22751

Wang, B., Lei, Y., Tian, S., Wang, T., Liu, Y., Patel, P., Jani, A.B., Mao, H., Curran, W. J., Liu, T., & Yang, X. (2019). Deeply supervised 3D fully convolutional networks with group dilated convolution for automatic MRI prostate segmentation. Medical Physics, 46(4), 1707–1718. https://doi.org/10.1002/mp.13416

Wang, Y., Midthun, D. E., Wampfler, J. A., Deng, B., Stoddard, S. M., Zhang, S., & Yang, P. (2015). Trends in the proportion of patients with lung cancer meeting screening criteria. JAMA: The Journal of the American Medical Association, 313(8), 853–855. https://doi.org/10.1001/jama.2015.413

Welch, H. G., & Albertsen, P. C. (2020). Reconsidering Prostate Cancer Mortality - The Future of PSA Screening. The New England Journal of Medicine, 382(16), 1557–1563. https://doi.org/10.1056/NEJMms1914228

Wesp, P., Grosu, S., Graser, A., Maurus, S., Schulz, C., Knösel, T., Fabritius, M. P., Schachtner, B., Yeh, B. M., Cyran, C. C., Ricke, J., Kazmierczak, P. M., & Ingrisch, M. (2022). Deep learning in CT colonography: differentiating premalignant from benign colorectal polyps. European Radiology, 32(7), 4749–4759. https://doi.org/10.1007/s00330-021-08532-2

Westphalen, A. C., McCulloch, C. E., Anaokar, J. M., Arora, S., Barashi, N. S., Barentsz, J. O., Bathala, T. K., Bittencourt, L. K., Booker, M. T., Braxton, V. G., Carroll, P. R., Casalino, D. D., Chang, S. D., Coakley, F. V., Dhatt, R., Eberhardt, S. C., Foster, B. R., Froemming, A. T., Fütterer, J. J., … Rosenkrantz, A. B. (2020). Variability of the Positive Predictive Value of PI-RADS for Prostate MRI across 26 Centers: Experience of the Society of Abdominal Radiology Prostate Cancer Diseasefocused Panel. Radiology, 296(1), 76–84. https://doi.org/10.1148/radiol.2020190646

Wing, P., & Langelier, M. H. (2009). Workforce shortages in breast imaging: impact on mammography utilization.AJR. American Journal of Roentgenology, 192(2), 370–378. https://doi.org/10.2214/AJR.08.1665

Winkel, D. J., Tong, A., Lou, B., Kamen, A., Comaniciu, D., Disselhorst, J. A., Rodríguez-Ruiz, A., Huisman, H., Szolar, D., Shabunin, I., Choi, M. H., Xing, P., Penzkofer, T., Grimm, R., von Busch, H., & Boll, D. T. (2021). A Novel Deep Learning Based Computer-Aided Diagnosis System Improves the Accuracy and Efficiency of Radiologists in Reading Biparametric Magnetic Resonance Images of the Prostate: Results of a Multireader, Multicase Study. Investigative Radiology, 56(10), 605–613. https://doi.org/10.1097/RLI.0000000000000780

World Health Organization. Regional Office for Europe. (2022). A short guide to cancer screening: increase effectiveness, maximize benefits and minimize harm. World Health Organization. Regional Office for Europe. https://apps.who.int/iris/handle/10665/351396

Zauber, A. G., Winawer, S. J., O’Brien, M. J., Lansdorp-Vogelaar, I., van Ballegooijen, M., Hankey, B. F., Shi, W., Bond, J. H., Schapiro, M., Panish, J. F., Stewart, E. T., & Waye, J. D. (2012). Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths. The New England Journal of Medicine, 366(8), 687–696. https://doi.org/10.1056/NEJMoa1100370

Zhang, B.-H., Yang, B.-H., & Tang, Z.-Y. (2004). Randomized controlled trial of screening for hepatocellular carcinoma. Journal of Cancer Research and Clinical Oncology, 130(7), 417–422.https://doi.org/10.1007/s00432-004-0552-0

Guide to Artificial Intelligence in Radiology

    Artificial intelligence (AI) is playing a growing role in all our lives and has shown promise in addressing some of the greatest current and upcoming societal challenges we face. The healthcare industry, though notoriously complex and resistant to disruption, potentially has a lot to gain from the use of AI. With an established history of leading digital transformation in healthcare and an urgent need for improved efficiency, radiology has been at the forefront of harnessing AI’s potential.

    This book covers how and why AI can address challenges faced by radiology departments, provides an overview of the fundamental concepts related to AI, and describes some of the most promising use cases for AI in radiology. In addition, the major challenges associated with the adoption of AI into routine radiological practice are discussed. The book also covers some crucial points radiology departments should keep in mind when deciding on which AI-based solutions to purchase. Finally, it provides an outlook on what new and evolving aspects of AI in radiology to expect in the near future.

    The healthcare industry has experienced a number of trends over the past few decades that demand a change in the way certain things are done. These trends are particularly salient in radiology, where the diagnostic quality of imaging scans has improved dramatically while scan times have decreased. As a result, the amount and complexity of medical imaging data acquired have increased substantially over the past few decades (Smith-Bindman et al., 2019; Winder et al., 2021) and are expected to continue to increase (Tsao, 2020). This issue is complicated by a widespread global shortage of radiologists (AAMC Report Reinforces Mounting Physician Shortage, 2021, Clinical Radiology UK Workforce Census 2019 Report, 2019). Healthcare workers, including radiologists, have an increasing workload (Bruls & Kwee, 2020; Levin et al., 2017) that contributes to burnout and medical errors (Harry et al., 2021). Being an essential service provider to virtually all other hospital departments, staff shortages within radiology have significant effects that spread throughout the hospital and to society as a whole (England & Improvement, 2019; Sutherland et al., n.d.).

    With an ageing global population and a rising burden of chronic illnesses, these issues are expected to pose even more of a challenge to the healthcare industry in the future.

    AI-based medical imaging solutions have the potential to ameliorate these challenges for several reasons. They are particularly suited to handling large, complex datasets (Alzubaidi et al., 2021). Moreover, they are well suited to automate some of the tasks traditionally performed by radiologists and radiographers, potentially freeing up time and making workflows within radiology departments more efficient (Allen et al., 2021; Baltruschat et al., 2021; Kalra et al., 2020; O’Neill et al., 2021; van Leeuwen et al., 2021; Wong et al., 2019). AI is also capable of detecting complex patterns in data that humans cannot necessarily find or quantify (Dance, 2021; Korteling et al., 2021; Kühl et al., 2020).

    The term “artificial intelligence” refers to the use of computer systems to solve specific problems in a way that simulates human reasoning. One fundamental characteristic of AI is that, like humans, these systems can tailor their solutions to changing circumstances. Note that, while these systems are meant to mimic on a fundamental level how humans think, their capacity to do so (e.g. in terms of the amount of data they can handle at one time, the nature and amount of patterns they can find in the data, and the speed at which they do so) often exceeds that of humans.

    AI solutions come in the form of computer algorithms, which are pieces of computer code representing instructions to be followed to solve a specific problem. In its most fundamental form, the algorithm takes data as an input, performs some computation on that data, and returns an output.

    An AI algorithm can be explicitly programmed to solve a specific task, analogous to a step-by-step recipe for baking a cake. On the other hand, the algorithm can be programmed to look for patterns within the data in order to solve the problem. These types of algorithms are known as machine learning algorithms. Thus, all machine learning algorithms are AI, but not all AI is machine learning. The patterns in the data that the algorithm can be explicitly programmed to look for or that it can “discover” by itself are known as features. An important characteristic of machine learning is that such algorithms learn from the data itself, and their performance improves the more data they are given.

    One of the most common uses of machine learning is in classification - assigning a piece of data a particular label. For example, a machine learning algorithm might be used to tell if a photo (the input) shows a dog or a cat (the label). The algorithm can learn to do so in a supervised or unsupervised way.

    Supervised learning

    In supervised learning, the machine learning algorithm is given data that has been labelled with the ground truth, in this example, photos of dogs and cats that have been labelled as such. The process then goes through the following phases:

    1.Training phase: The algorithm learns the features associated with dogs and cats using the aforementioned data (training data).
    2.Test phase: The algorithm is then given a new set of photos (the test data), it labels them and the performance of the algorithm on that data is assessed.

    In some cases, there is a phase in between training and test, known as the validation phase. In this phase, the algorithm is given a new set of photos (not included in either the training or test data), its performance is assessed on this data, and the model is tweaked and retrained on the training data. This is repeated until some predefined performance-based criterion is reached, and the algorithm then enters the test phase.

    Unsupervised learning

    In unsupervised learning, the algorithm identifies features within the input data that allow it to assign classes to the individual data points without being told explicitly what those classes are or should be. Such algorithms can identify patterns or group data points together without human intervention and include clustering and dimensionality reduction algorithms. Not all machine learning algorithms perform classification. Some are used to predict a continuous metric (e.g. the temperature in four weeks’ time) instead of a discrete label (e.g. cats vs dogs). These are known as regression algorithms.

    Neural networks and deep learning

    A neural network is made up of an input layer and an output layer, which are themselves composed of nodes. In simple neural networks, features that are manually derived from a dataset are fed into the input layer, which performs some computations, the results of which are relayed to the output layer. In deep learning, multiple “hidden” layers exist between the input and output layers. Each node of the hidden layers performs calculations using certain weights and relays the output to the next hidden layer until the output layer is reached.

    In the beginning, random values are assigned to the weights and the accuracy of the algorithm is calculated. The values of the weights are then iteratively adjusted until a set of weight values that maximize accuracy is found. This iterative adjustment of the weight values is usually done by moving backwards from the output layer to the input layer, a technique called backpropagation. This entire process is done on the training data.

    Performance evaluation

    Understanding how the performance of AI algorithms is assessed is key to interpreting the AI literature. Several performance metrics exist for assessing how well a model performs certain tasks. No single metric is perfect, so a combination of several metrics provides a fuller picture of model performance.

    In regression, the most commonly used metrics include:

    • Mean absolute error (MAE): the average difference between the predicted values and the ground truth.
    • Root mean square error (RMSE): the differences between the predicted values and the ground truth are squared and then averaged over the sample. Then the square root of the average is taken. Unlike the MAE, the RMSE thus gives higher weight to larger differences.
    • R2: the proportion of the total variance in the ground truth explained by the variance in the predicted values. It ranges from 0 to 1.

    The following metrics are commonly used in classification tasks:

    • Accuracy: this is the proportion of all predictions that were predicted correctly. It ranges from 0 to 1.
    • Sensitivity: also known as the true positive rate (TPR) or recall, this is the proportion of true positives that were predicted correctly. It ranges from 0 to 1.
    • Specificity: Also known as the true negative rate (TNR), this is the proportion of true negatives that were predicted correctly. It ranges from 0 to 1.
    • Precision: also known as positive predictive value (PPV), this is the proportion of positive classifications that were predicted correctly. It ranges from 0 to 1.

    An inherent trade-off exists between sensitivity and specificity. The relevant importance of each, as well as their interpretation, highly depends on the specific research question and classification task.

    Importantly, although classification models are meant to reach a binary conclusion, they are inherently probability-based. This means that these models will output a probability that a data point belongs to one class or another. In order to reach a conclusion on the most likely class, a threshold is used. Metrics such as accuracy, sensitivity, specificity and precision refer to the performance of the algorithm based on a certain threshold. The area under the receiver operating characteristic curve (AUC) is a threshold-independent performance metric. The AUC can be interpreted as the probability that a random positive example is ranked higher by the algorithm than a random negative example.

    In image segmentation tasks, which are a type of classification task, the following metrics are commonly used:

    • Dice similarity coefficient: a measure of overlap between two sets (e.g. two images) that is calculated as two times the number of elements common to the sets divided by the sum of the number of elements in each set. It ranges from 0 (no overlap) to 1 (perfect overlap).
    • Hausdorff distance: a measure of how far two sets (e.g. two images) within a space are far from each other. It is basically the largest distance from one point in one set to the closest point in the other set.

    Internal and external validity

    Internally valid models perform well in their task on the data being used to train and validate them. The degree to which they are internally valid is assessed using the performance metrics outlined above and depends on the characteristics of the model itself and the quality of the data that the model was trained and validated on.

    Externally valid models perform well in their tasks on new data (Ramspek et al., 2021). The better the model performs on data that differs from the data the models were trained and validated on, the higher the external validity. In practice, this often requires the performance of the models to be tested on data from hospitals or geographical areas that were not part of the model’s training and validation datasets.

    Guidelines for evaluating AI research

    Several guidelines have been developed to assess the evidence behind AI-based interventions in healthcare (X. Liu et al., 2020; Mongan et al., 2020; Shelmerdine et al., 2021; Weikert et al., 2021). These provide a template for those doing AI research in healthcare and ensure that relevant information is reported transparently and comprehensively, but can also be used by other stakeholders to assess the quality of published research. This helps ensure that AI-based solutions with substantial potential or actual limitations, particularly those caused by poor reporting (Bozkurt et al., 2020; D. W. Kim et al., 2019; X. Liu et al., 2019; Nagendran et al., 2020; Yusuf et al., 2020), are not prematurely adopted (CONSORT-AI and SPIRIT-AI Steering Group, 2019). Guidelines have also been proposed for evaluating the trustworthiness of AI-based solutions in terms of transparency, confidentiality, security, and accountability (Buruk et al., 2020; Lekadir et al., 2021; Zicari et al., 2021).

    Over the past few years, AI has shown great potential in addressing a broad range of tasks within a medical imaging department, including many that happen before the patient is scanned. Implementations of AI to improve the efficiency of radiology workflows prior to patient scanning are sometimes referred to as “upstream AI” (Kapoor et al., 2020; M. L. Richardson et al., 2021).

    Scheduling

    One promising upstream AI application is predicting whichpatients arelikelytomisstheirscanappointments. Missed appointments are associated with significantly increased workload and costs (Dantas et al., 2018). Using a Gradient Boosting approach, Nelson et al. predicted missed hospital magnetic resonance imaging (MRI) appointments in the United Kingdom’s National Health Service (NHS) with high accuracy (Nelson et al., 2019). Their simulations also suggested that acting on the predictions of this model by targeting patients who are likely to miss their appointments would potentially yield a net benefit of several pounds per appointment across a range of model thresholds and missed appointment rates (Nelson et al., 2019). Similar results were recently found in a study of a single hospital in Singapore. For the 6-month period following the deployment of the predictive tool they were able to significantly reduce the no show rate from 19.3 % tp 15.9 % which translated into a potential economic benefit of $180,000 (Chong et. al., 2020).

    Scheduling scans in a radiology department is a challenging endeavour because, although it is largely an administrative task, it depends heavily on medical information. The task of assigning patients to specific appointments thus often requires the input of someone with domain knowledge, which stipulates that either the person making the appointments must be a radiologist or radiology technician, or these people will have to provide input regularly. In either scenario, the process is somewhat inefficient and can potentially be streamlined using AI-based algorithms that check scan indications and contraindications and provide the people scheduling the scans with information about scan urgency (Letourneau-Guillon et al., 2020).

    Protocolling

    Depending on hospital or clinic policy, the decision on what exact scan protocol a patient receives is usually made based on the information on the referring physician’s scan request and the judgement of the radiologist. This is often supplemented by direct communication between the referring physician and radiologist and the radiologist’s review of the patient’s medical information. This process improves patient care (Boland et al., 2014) but can be time-consuming and inefficient, particularly with modalities like MRI, where a large number of protocol permutations exist. In one study, protocolling alone accounted for about 6 % of the radiologist’s working time (Schemmel et al., 2016). Radiologists are also often interrupted by tasks such as protocolling when interpreting images, despite the fact that the latter is considered a radiologist’s primary responsibility (Balint et al., 2014; J.-P. J. Yu et al., 2014).

    Interpretation of the narrative text of the referring physician’s scan request has been attempted using natural language classifiers, the same technology used in chatbots and virtual assistants. Natural language classifiers based on deep learning have shown promise in assigning patients to either a contrast-enhanced or non-enhanced MRI protocol for musculoskeletal MRI, with an accuracy of 83 % (Trivedi et al., 2018) and 94 % (Y. H. Lee, 2018). Similar algorithms have shown an accuracy of 95 % for predicting the appropriate brain MRI protocol using a combination of up to 41 different MRI sequences (Brown & Marotta, 2018). Across a wide range of body regions, a deep-learning-based natural language classifier decided based on the narrative text of the scan requests whether to automatically assign a specific computed tomography (CT) or MRI protocol (which it did with 95 % accuracy) or, in more difficult cases, recommend a list of three most appropriate protocols to the radiologist (which it did with 92 % accuracy) (Kalra et al., 2020).

    AI has also been used to decide whether already protocolled scans need to be extended, a decision which has to be made in real-time while the patient is inside the scanner. One such example is in prostate MRI, where a decision on whether to administer a contrast agent is often made after the non-contrast sequences. Hötker et al. found that a convolutional neural network (CNN) assigned 78 % of patients to the appropriate prostate MRI protocol (Hötker et al., 2021). The sensitivity of the CNN for the need for contrast was 94.4 % with a specificity of 68.8 % and only 2 % of patients in their study would have had to be called back for a contrast- enhanced scan (Hötker et al., 2021).

    Image quality improvement and monitoring

    Many AI-based solutions that work in the background of radiology workflows to improve image quality have recently been established. These include solutions for monitoring image quality, reducing image artefacts, improving spatial resolution, and speeding up scans.

    Such solutions are entering the radiology mainstream, particularly for computed tomography, which for decades used established but artefact-prone methods for reconstructing interpretable images from the raw sensor data (Deák et al., 2013; Singh et al., 2010).

    These are gradually being replaced by deep-learning- based reconstruction methods, which improve image quality while maintaining low radiation doses (Akagi et al., 2019; H. Chen et al., 2017; Choe et al., 2019; Shan et al., 2019). This reconstruction is performed on supercomputers on the CT scanner itself or on the cloud. The balance between radiation dose and image quality can be adjusted on a protocol-specific basis to tailor scans to individual patients and clinical scenarios (McLeavy et al., 2021; Willemink & Noël, 2019). Such approaches have found particular use when scanning children, pregnant women, and obese patients as well as CT scans of the urinary tract and heart (McLeavy et al., 2021).

    AI-based solutions have also been used to speed up scans while maintaining diagnostic quality. Scan time reduction not only improves overall efficiency but also contributes to an overall better patient experience and compliance with imaging examination. A multi- centre study of spine MRI showed that a deep-learning- based image reconstruction algorithm that enhanced images using filtering and detail-preserving noise reduction reduced scan times by 40 % (Bash, Johnson, et al., 2021). For T1-weighted MRI scans of the brain, a similar algorithm that improves image sharpness and reduces image noise reduced scan times by 60 % while maintaining the accuracy of brain region volumetry compared to standard scans (Bash, Wang, et al., 2021).

    In routine radiological practice, images often contain artefacts that reduce their interpretability. These artefacts are the result of characteristics of the specific imaging modality or protocol used or factors intrinsic to the patient being scanned, such as the presence of foreign bodies or the patient moving during the scan. Particularly with MRI, imaging protocols that demand fast scanning often introduce certain artefacts to the reconstructed image. In one study, a deep-learning- based algorithm reduced banding artefacts associated with balanced steady-state free precession MRI sequences of the brain and knee (K. H. Kim & Park, 2017). For real-time imaging of the heart using MRI, another study found that the aliasing artefacts introduced by the data undersampling were reduced by using a deep-learning-based approach (Hauptmann et al., 2019). The presence of metallic foreign bodies such as dental, orthopaedic or vascular implants is a common patient-related factor causing image artefacts in both CT and MRI (Boas & Fleischmann, 2012; Hargreaves et al., 2011). Although not yet well established, several deep-learning-based approaches for reducing these artefacts have been investigated (Ghani & Clem Karl, 2019; Puvanasunthararajah et al., 2021; Zhang & Yu, 2018). Similar approaches are being tested for reducing motion-related artefacts in MRI (Tamada et al., 2020; B. Zhao et al., 2022).

    AI-based solutions for monitoring image quality potentially reduce the need to call patients back to repeat imaging examinations, which is a common problem (Schreiber-Zinaman & Rosenkrantz, 2017). A deep-learning-based algorithm that identifies the radiographic view acquired and extracts quality-related metrics from ankle radiographs was able to predict image quality with about 94 % accuracy (Mairhöfer et al., 2021). Another deep-learning-based approach was capable of predicting nondiagnostic liver MRI scans with a negative predictive value of between 86 % and 94 % (Esses et al., 2018). This real-time automated quality control potentially allows radiology technicians to rerun scans or run additional scans with greater diagnostic value.

    Scan reading prioritization

    With staff shortages and increasing scan numbers, radiologists face long reading lists. To optimize efficiency and patient care, AI-based solutions have been suggested as a way to prioritize which scans radiologists read and report first, usually by screening acquired images for findings that require urgent intervention (O’Connor & Bhalla, 2021). This has been most extensively studied in neuroradiology, where moving CT scans that were found to have intracranial haemorrhage by an AI-based tool to the top of the reading list reduced the time it took radiologists to view the scans by several minutes (O’Neill et al., 2021). Another study found that the time-to diagnosis (which includes the time from image acquisition to viewing by the radiologist and the time to read and report the scans) was reduced from 512 to 19 minutes in an outpatient setting when such a worklist prioritization was used (Arbabshirani et al., 2018). A simulation study using AI-based worklist prioritization based on identifying urgent findings on chest radiographs (such as pneumothorax, pleural effusions, and foreign bodies) also found a substantial reduction in the time it took to view and report the scans compared to standard workflow prioritization (Baltruschat et al., 2021).

    Image interpretation

    Currently, the majority of commercially available AI- based solutions in medical imaging focus on some aspect of analyzing and interpreting images (Rezazade Mehrizi et al., 2021; van Leeuwen et al., 2021). This includes segmenting parts of the image (for surgical or radiation therapy targeting, for example), bringing suspicious areas to radiologists’ attention, extracting imaging biomarkers (radiomics), comparing images across time, and reaching specific imaging diagnoses.

    Neurology

    ¡ 29–38 % of commercially available AI-based applications in radiology (Rezazade Mehrizi et al., 2021; van Leeuwen et al., 2021).

    Most commercially available AI-based solutions targeted at neuroimaging data aim to detect and characterize ischemic stroke, intracranial haemorrhage, dementia, and multiple sclerosis (Olthof et al., 2020). Several studies have shown excellent accuracy of AI- based methods for the detection and classification of intraparenchymal, subarachnoid, and subdural haemorrhage on head CT (Flanders et al., 2020; Ker et al., 2019; Kuo et al., 2019). Subsequent studies showed that, compared to radiologists, some AI-based solutions have substantially lower false positive and negative rates (Ginat, 2020; Rao et al., 2021). In ischemic stroke, AI-based solutions have largely focused on the quantification of the infarct core (Goebel et al., 2018; Maegerlein et al., 2019), the detection of large vessel occlusion (Matsoukas et al., 2022; Morey et al., 2021; Murray et al., 2020; Shlobin et al., 2022), and the prediction of stroke outcomes (Bacchi et al., 2020; Nielsen et al., 2018; Y. Yu et al., 2020, 2021).

    In multiple sclerosis, AI has been used to identify and segment lesions (Nair et al., 2020; S.-H. Wang et al., 2018), which can be particularly helpful for the longitudinal follow-up of patients. It has also been used to extract imaging features associated with progressive disease and conversion from clinically isolated syndrome to definite multiple sclerosis (Narayana et al., 2020; Yoo et al., 2019). Other applications of AI in neuroradiology include the detection of intracranial aneurysms (Faron et al., 2020; Nakao et al., 2018; Ueda et al., 2019) and the segmentation of brain tumours (Kao et al., 2019; Mlynarski et al., 2019; Zhou et al., 2020) as well as the prediction of brain tumour genetic markers from imaging data (Choi et al., 2019; J. Zhao et al., 2020)

    Chest

    ¡ 24 %–31 % of commercially available AI-based applications in radiology (Rezazade Mehrizi et al., 2021; van Leeuwen et al., 2021).

    When interpreting chest radiographs, radiologists detected substantially more critical and urgent findings when aided by a deep-learning-based algorithm, and did so much faster than without the algorithm (Nam et al., 2021). Deep-learning-based image interpretation algorithms have also been found to improve radiology residents’ sensitivity for detecting urgent findings on chest radiographs from 66 % to 73 % (E. J. Hwang, Nam, et al., 2019). Another study which focused on a broader range of findings on chest radiographs also found that radiologists aided by a deep-learning-based algorithm had higher diagnostic accuracy than radiologists who read the radiographs without assistance (Seah et al., 2021). The uses of AI in chest radiology also extend to cross-sectional imaging like CT. A deep learning algorithm was found to detect pulmonary embolism on CT scans with high accuracy (AUC = 0.85) (Huang, Kothari, et al., 2020). Moreover, a deep learning algorithm was 90 % accurate in detecting aortic dissection on non-contrast-enhanced CT scans, similar to the performance of radiologists (Hata et al., 2021).

    Outside the emergency setting, AI-based solutions have been widely tested and implemented for tuberculosis screening on chest radiographs (E. J. Hwang, Park, et al., 2019; S. Hwang et al., 2016; Khan et al., 2020; Qin et al., 2019; WHO Operational Handbook on Tuberculosis Module 2: Screening – Systematic Screening for Tuberculosis Disease, n.d.). In addition, they have been useful for lung cancer screening both in terms of detecting lung nodules on CT (Setio et al., 2017) and chest radiographs (Li et al., 2020) and by classifying whether nodules are likely to be malignant or benign (Ardila et al., 2019; Bonavita et al., 2020; Ciompi et al., 2017; B. Wu et al., 2018). AI-based solutions also show great promise for the diagnosis of pneumonia, chronic obstructive pulmonary disease, and interstitial lung disease (F. Liu et al., 2021).

    Breast

    ¡ 11 % of commercially available AI-based applications in radiology (Rezazade Mehrizi et al., 2021; van Leeuwen et al., 2021).

    So far, many of the AI-based algorithms targeting breast imaging aim to reduce the workload of radiologists reading mammograms. Ways to do this have included using AI-based algorithms to triage out negative mammograms, which in one study was associated with a reduction in radiologists’ workload by almost one-fifth (Yala et al., 2019). Other studies that have replaced second readers of mammograms with AI- based algorithms have shown that this leads to fewer false positives and false negatives as well as reduces the workload of the second reader by 88 % (McKinney et al., 2020).

    AI-based solutions for mammography have also been found to increase the diagnostic accuracy of radiologists (McKinney et al., 2020; Rodríguez-Ruiz et al., 2019; Watanabe et al., 2019) and some have been found to be highly accurate in independently detecting and classifying breast lesions (Agnes et al., 2019; Al- Antari et al., 2020; Rodriguez-Ruiz et al., 2019).
    Despite this, a recent systematic review of 36 AI- based algorithms found that these studies were of poor methodological quality and that all algorithms were less accurate than the consensus of two or more radiologists (Freeman et al., 2021). AI-based algorithms have nonetheless shown potential for extracting cancer-predictive features from mammograms beyond mammographic breast density (Arefan et al., 2020; Dembrower et al., 2020; Hinton et al., 2019). Beyond mammography, AI-based solutions have been developed for detecting and classifying breast lesions on ultrasound (Akkus et al., 2019; Park et al., 2019; G.- G. Wu et al., 2019) and MRI (Herent et al., 2019).

    Cardiac

    ¡ 11 % of commercially available AI-based applications in radiology (Rezazade Mehrizi et al., 2021; van Leeuwen et al., 2021).

    Cardiac radiology has always been particularly challenging because of the difficulties inherent in acquiring images of a constantly moving organ. Because of this, it has benefited immensely from advances in imaging technology and seems set to benefit greatly from AI as well (Sermesant et al., 2021). Most of the AI-based applications of the cardiovascular system use MRI, CT or ultrasound data (Weikert et al., 2021). Prominent examples include the automated calculation of ejection fraction on echocardiography, quantification of coronary artery calcification on cardiac CT, determination of right ventricular volume on CT pulmonary angiography, and determination of heart chamber size and thickness on cardiac MRI (Medical AI Evaluation, n.d., The Medical Futurist, n.d.). AI-based solutions for the prediction of patients likely to respond favourably to cardiac interventions, such as cardiac resynchronization therapy, based on imaging and clinical parameters have also shown great promise (Cikes et al., 2019; Hu et al., 2019). Changes in cardiac MRI not readily visible to human readers but potentially useful for differentiating different types of cardiomyopathies can also be detected using AI through texture analysis (Neisius et al., 2019; J. Wang et al., 2020) and other radiomic approaches (Mancio et al., 2022).

    Musculoskeletal

    ¡ 7–11 % of commercially available AI-based applications in radiology (Rezazade Mehrizi et al., 2021; van Leeuwen et al., 2021).

    Promising applications of AI in the assessment of muscles, bones and joints include applications where human readers generally show poor between- and within-rater reliability, such as the determination of skeletal age based on bone radiographs (Halabi et al., 2019; Thodberg et al., 2009) and screening for osteoporosis on radiographs (Kathirvelu et al., 2019; J.-S. Lee et al., 2019) and CT (Pan et al., 2020). AI- based solutions have also shown promise for detecting fractures on radiographs and CT (Lindsey et al., 2018; Olczak et al., 2017; Urakawa et al., 2019). One systematic review of AI-based solutions for fracture detection in several different body parts showed AUCs ranging from 0.94 to 1.00 and accuracies of 77 % to 98 % (Langerhuizen et al., 2019). AI-based solutions have also achieved accuracies similar to radiologists for classification of the severity of degenerative changes of the spine (Jamaludin et al., 2017) and extremity joints (F. Liu et al., 2018; Thomas et al., 2020). AI-based solutions have also been developed to determine the origin of skeletal metastases (Lang et al., 2019) and the classification of primary bone tumours (Do et al., 2017).

    Abdomen and pelvis

    ¡ 4 % of commercially available AI-based applications in radiology (Rezazade Mehrizi et al., 2021; van Leeuwen et al., 2021).

    Much of the efforts in using AI in abdominal imaging have thus far concentrated on the automated segmentation of organs such as the liver (Dou et al., 2017), spleen (Moon et al., 2019), pancreas (Oktay et al., 2018), and kidneys (Sharma et al., 2017). In addition, a systematic review of 11 studies using deep learning for the detection of malignant liver masses showed accuracies of up to 97 % and AUCs of up to 0.92 (Azer, 2019).

    Other applications of AI in abdominal radiology include the detection of liver fibrosis (He et al., 2019; Yasaka et al., 2018), fatty liver disease, hepatic iron content, the detection of free abdominal gas on CT, and automated volumetry and segmentation of the prostate (AI for Radiology, n.d.).

    Despite the great potential of AI in medical imaging, it has yet to find widespread implementation and impact in routine clinical practice. This research-to- clinic translation is being hindered by several complex and interrelated issues that directly or indirectly lower the likelihood of AI-based solutions being adopted. One major way they do so is by creating a lack of trust in AI- based solutions by key stakeholders such as regulators, healthcare professionals and patients (Cadario et al., 2021; Esmaeilzadeh, 2020; J. P. Richardson et al., 2021; Tucci et al., 2022).

    Generalizability

    One major challenge is to develop AI-based solutions that continue to perform well in new, real-world scenarios. In a large systematic review, almost half of the studied AI-based medical imaging algorithms reported a greater than 0.05 decrease in the AUC when tested on new data (A. C. Yu et al., 2022). This lack of generalizability can lead to adverse effects on how well the model performs in a real-world scenario.

    If a solution performs poorly when tested on a dataset with a similar or identical distribution to the training dataset, it is said to lack narrow generalizability and is often a consequence of overfitting (Eche et al., 2021). Potential solutions for overfitting are using larger training datasets and reducing the model’s complexity. If a solution performs poorly when tested on a dataset with a different distribution to the training dataset (e.g. a different distribution of patient ethnicities), it is said to lack broad generalizability (Eche et al., 2021). Solutions to poor broad generalizability include stress-testing the model on datasets with different distributions from the training dataset (Eche et al., 2021).

    AI solutions are often developed in a high-resource environment such as large technology companies and academic medical centres in wealthy countries. It is likely that findings and performance in these high-resource contexts will fail to generalize to lower- resource contexts such as smaller hospitals, rural areas or poorer countries (Price & Nicholson, 2019), which complicates the issue further.

    Risk of bias

    Biases can arise in AI-based solutions due to data or human factors. The former occurs when the data used to train the AI solution does not adequately represent the target population. Datasets can be unrepresentative when they are too small or have been collected in a way that misrepresents a certain population category. AI solutions trained on unrepresentative data perpetuate biases and perform poorly in the population categories underrepresented or misrepresented in the training data. The presence of such biases has been empirically shown in many AI-based medical imaging studies (Larrazabal et al., 2020; Seyyed-Kalantari et al., 2021).

    AI-based solutions are prone to several subjective and sometimes implicitly or explicitly prejudiced decisions during their development by humans. These human factors include how the training data is selected, how it is labelled, and how the decision is made to focus on the specific problem the AI-based solution intends to solve (Norori et al., 2021). Some recommendations and tools are available to help minimize the risk of bias in AI research (AIF360: A Comprehensive Set of Fairness Metrics for Datasets and Machine Learning Models, Explanations for These Metrics, and Algorithms to Mitigate Bias in Datasets and Models, n.d., IBM Watson Studio - Model Risk Management, n.d.; Silberg & Manyika, 2019).

    Data quantity, quality and variety

    Problems such as bias and lack of generalizability can be mitigated by ensuring that training data is of sufficient quantity, quality and variety. However, this is difficult to do because patients are often reluctant to share their data for commercial purposes (Aggarwal, Farag, et al., 2021; Ghafur et al., 2020; Trinidad et al., 2020), hospitals and clinics are usually not equipped to make this data available in a useable and secure manner, and organizing and labelling the data is time- consuming and expensive.

    Many datasets can be used for a number of different purposes, and sharing data between companies can help make the process of data collection and organization more efficient, as well as increase the amount of data available for each application. However, developers are often reluctant to share data with each other, or even reveal the exact source of their data, to stay competitive.

    Data protection and privacy

    The development and implementation of AI-based solutions require that patients are explicitly informed about, and give their consent to, the use of their data for a particular purpose and by certain people. This data also has to be adequately protected from data breaches and misuse. Failure to ensure this greatly undermines the public’s trust in AI-based solutions and hinders their adoption. While regulations governing health data privacy state that the collection of fully anonymized data does not require explicit patient consent (General Data Protection Regulation (GDPR) – Official Legal Text, 2016; Office for Civil Rights (OCR), 2012) and in theory protects from the data being misused, whether or not imaging data can be fully anonymized is controversial (Lotan et al., 2020; Murdoch, 2021). Whether consent can be truly informed considering the complexity of the data being acquired, and the resulting myriad of potential future uses of the data, is also disputed (Vayena & Blasimme, 2017).

    IT infrastructure

    Among hospital departments, radiology has always been at the forefront ofdigitalization. AI-based solutions that focus on image processing and interpretation are likely to find the prerequisite infrastructure in most radiology departments, for example for linking imaging equipment to computers for analysis and for archiving images and other outputs. However, most radiology departments are likely to require significant infrastructure upgrades for other applications of AI, particularly those requiring the integration of information from multiple sources and having complex outputs. Moreover, it is important to keep in mind that the distribution of necessary infrastructure is highly unequal across and within countries (Health Ethics & Governance, 2021).

    In terms of computing power, radiology departments will either have to invest resources into the hardware and personnel necessary to run these AI-based solutions or opt for cloud-based solutions. The former comes with an extra cost but allows data processing within the confines of the hospital or clinic’s local network. Cloud-based solutions for computing (known as “infrastructure as a service” or “IaaS”) are often considered the less secure and less trustworthy option, but this depends on a number of factors and is thus not always true (Baccianella & Gough, n.d.). Guidelines on what to consider when procuring cloud-based solutions in healthcare are available (Cloud Security for Healthcare Services, 2021).

    Lack of standardization, interoperability, and integrability

    The problem of infrastructure becomes even more complicated when considering how fragmented the AI medical imaging market currently is (Alexander et al., 2020). It is therefore likely that in the near future a single department will have several dozen AI-based solutions from different vendors running simultaneously. Having a separate self-contained infrastructure (e.g. a workstation or server) for each of these would be incredibly complicated and difficult to manage. Suggested solutions for this have included AI solution “marketplaces”, similar to app stores (Advanced AI Solutions for Radiology, n.d., Curated Marketplace, 2018, Imaging AI Marketplace - Overview, n.d., Sectra Amplifier Marketplace, 2021, The Nuance AI Marketplace for Diagnostic Imaging, n.d.), and development of an overarching vendor-neutral infrastructure (Leiner et al., 2021). The successful implementation of such solutions requires close partnerships between AI solution developers, imaging vendors and information technology companies.

    Interpretability

    It is often impossible to understand exactly how AI- based solutions come to their conclusions, particularly with complex approaches like deep learning. This reduces how transparent the decision-making process for procuring and approving these solutions can be, makes the identification of biases difficult, and makes it harder for clinicians to explain the outputs of these solutions to their patients and to determine whether a solution is working properly or has malfunctioned (Char et al., 2018; Reddy et al., 2020; Vayena et al., 2018; Whittlestone et al., 2019). Some have suggested that techniques that help humans understand how AI- based algorithms made certain decisions or predictions (“interpretable” or “explainable” AI) might help mitigate these challenges. However, others have argued that currently available techniques are unsuitable for understanding individual decisions of an algorithm and have warned against relying on them for ensuring that algorithms work in a safe and reliable way (Ghassemi et al., 2021).

    Liability

    In healthcare systems, a framework of accountability ensures that healthcare workers and medical institutions can be held responsible for adverse effects resulting from their actions. The question of who should be held accountable for the failures of an AI- based solution is complicated. For pharmaceuticals, for example, the accountability for inherent failures in the product or its use often lies with either the manufacturer or the prescriber. One key difference is that AI-based systems are continuously evolving and learning, and so inherently work in a way that is independent of what their developers could have foreseen (Yeung, 2018). To the end-user such as the healthcare worker, the AI- based solution may be opaque and so they may not be able to tell when the solution is malfunctioning or inaccurate (Habli et al., 2020; Yeung, 2018).

    Brittleness

    Despite substantial progress in their development over the past few years, deep learning algorithms are still surprising brittle. This means that, when the algorithm faces a scenario that differs substantially from what it faced during training, it cannot contextualize and often produces nonsensical or inaccurate results. This happens because, unlike humans, most algorithms learn to perceive things within the confines of certain assumptions, but fail to generalize outside these assumptions. As an example of how this can be abused with malicious intent, subtle changes to medical images, imperceptible by humans, can render the results of disease-classifying algorithms inaccurate (Finlayson et al., 2018). The lack of interpretability of many AI-based solutions compounds this problem because it makes it difficult to troubleshoot how they reached the wrong conclusion.

    So far, more than 100 AI-based products have gained conformité européenne (CE) marking or Food and Drug Adminstration (FDA) clearance. These products can be found in continuously updated and searchable online databases curated by the FDA (Center for Devices & Radiological Health, n.d.), the American College of Radiology (Assess-AI, n.d.), and others (AI for Radiology, n.d., The Medical Futurist, n.d.; E. Wu et al., 2021). The increasing number of available products, the inherent complexity of many of these solutions, and the fact that many people who usually make purchasing decisions in hospitals are not familiar with evaluating such products make it important to think carefully when deciding on which product to purchase. Such decisions will need to be made after incorporating input from healthcare workers, information technology (IT) professionals, as well as management, finance, legal, and human resources professionals within hospitals.

    Deciding on whether to purchase an AI-based solution in radiology, as well as which of the increasing number of commercially available solutions to purchase, includes considerations of quality, safety, and finances. Over the past few years, several guidelines have emerged to help potential buyers make these decisions (A Buyer’s Guide to AI in Health and Care, 2020; Omoumi et al., 2021; Reddy et al., 2021), and these guidelines are likely to evolve in the future with changing expectations from customers, regulatory bodies, and stakeholders involved in reimbursement decisions.

    First of all, it has to be clear to the potential buyer what the problem is and whether AI is the appropriate approach to this solution, or whether alternatives exist that are more advantageous on balance. If AI is the appropriate approach, buyers should know exactly what a potential AI-based product’s scope of the solution is - i.e. what specific problem the AI-based solution is designed to solve and in what specific circumstances. This includes whether the solution is intended for screening, diagnosis, monitoring, treatment recommendation or another application. It also includes the intended users of the solution and what kind of specific qualifications or training they are expected to have in order to be able to operate the solution and interpret its outputs. It needs to be clear to buyers whether the solution is intended to replace certain tasks that would normally be performed by the end-user, act as a double-reader, as a triaging mechanism, or for other tasks like quality control. Buyers should also understand whether the solution is intended to provide “new” information (i.e. information that would otherwise be unavailable to the user without the solution), improve the performance of an existing task beyond a human’s or other non-AI-based solution’s performance or if it is intended to save time or other resources.

    Buyers should also have access to information that allows them to assess the potential benefits of the AI solution, and this should be backed up by published scientific evidence for the efficacy and cost-efficiency of the solution. How this is done will depend highly on the solution itself and the context in which it is expected to be deployed, but guidelines for this are available (National Institute for Health and Care Excellence (NICE), n.d.). Some questions to ask here would be: How much of an influence will the solution have on patient management? Will it improve diagnostic performance? Will it save time and money? Will it affect patients’ quality of life? It should also be clear to the buyer who exactly is expected to benefit from the use of this solution (Radiologists? Clinicians? Patients? The healthcare system or society as a whole?).

    As with any healthcare intervention, all AI-based solutions come with potential risks, and these should be made clear to the buyer. Some of these risks might have legal consequences, such as the potential for misdiagnosis. These risks should be quantified, and potential buyers should have a framework for dealing with them, including identifying a framework for accountability within the organizations implementing these solutions. Buyers should also ensure they clearly understand the potential negative effects on radiologists’ training and the potential disruption to radiologists’ workflows associated with the use of these solutions.

    Specifics of the AI solution’s design are also relevant to the decision on whether or not to purchase it. These include how robust the solution is to differences between vendors and scanning parameters, the circumstances under which the algorithm was trained (including potential confounding factors), and the way that performance was assessed. It should also be clear to buyers if and how potential sources of bias were accounted for during development. Because a core characteristic of AI-based solutions is their ability to continuously learn from new data, whether and how exactly this retraining is incorporated into the solution with time should also be clear to the buyer, including whether or not new regulatory approval is needed with each iteration. This also includes whether or not retraining is required, for example, due to changes in imaging equipment at the buyer’s institution.

    The main selling points of many AI-based solutions are ease-of-use and improved workflows. Therefore, potential buyers should carefully scrutinize how these solutions are to be integrated into existing workflows, including inter-operability with PACS and electronic medical record systems. Whether or not the solution requires extra hardware (e.g. graphical processing units) or software (e.g. for visualization of the solution’s outputs), or if it can readily be integrated into the existing information technology infrastructure of the buyer’s organization influences the overall cost of the solution for the buyer and is therefore also a critical consideration. In addition, the degree of manual interaction required, both under normal circumstances and for troubleshooting, should be known to the buyer. All potential users of the AI solution should be involved in the purchasing process to ensure that they are familiar with it and that it meets their professional ethical standards and suits their needs.

    From a regulatory perspective, it should be clear to the buyer whether the solution complies with medical device and data protection regulations. Has the solution been approved in the buyer’s country? If so, under which risk classification? Buyers should also consider creating data flow maps that display how the data flows in the operation of the AI-based solution, including who has access to the data.

    Finally, there are other factors to consider which are not necessarily unique to AI-based solutions and which buyers might be familiar with from purchasing other types of solutions. This includes the licensing model of the solution, how users are to be trained on using the solution, how the solution is maintained, how failures in the solution are dealt with, and whether additional costs are to be expected when scaling up the solution’s implementation (e.g. using the solution for more imaging equipment or more users). This allows the potential buyer to anticipate the current and future costs of purchasing the solution.

    The past decade of increasing interest and progress in AI-based solutions for medical imaging has set the stage for a number of trends that are likely to appear or intensify in the near future.

    Firstly, there is an increasing sentiment that, although AI holds a great deal of promise for interpretive applications (such as the detection of pathology), non-interpretive AI-based solutions might hold the most potential in terms of instilling efficiency into radiology workflows and improving patient experiences. This trend towards involving AI earlier in the patient management process is likely to extend to AI increasingly acting as a clinical decision support system to guide when and which imaging scans are performed.

    For this to happen, AI needs to be integrated into existing clinical information systems, and the specific algorithms used need to be able to handle more varied data. This will likely pave the way for the development of algorithms that are capable of integrating demographic, clinical, and laboratory patient data to make recommendations about patient management (Huang, Pareek, et al., 2020; Rockenbach, 2021). The previously mentioned natural language processing algorithms that have been used to interpret scan requests may be useful candidates for this.

    In addition, we are likely to see AI algorithms that can interpret multiple different types of imaging data from the same patient. Currently, less than 5 % of commercially available AI-based solutions in medical imaging work with more than one imaging modality (Rezazade Mehrizi et al., 2021; van Leeuwen et al., 2021) despite the fact that the typical patient in a hospital receives multiple imaging scans during their stay (Shinagare et al., 2014). With this, it is also likely that more AI-based solutions will be developed that target hitherto neglected modalities such as nuclear imaging techniques and ultrasound.

    The current market for AI-based solutions in radiology is spread across a relatively large number of companies (Alexander et al., 2020). Potential users are likely to expect a streamlined integration of these products in their workflows, which can be challenging in such a fragmented market. Improved integration can be achieved in several different ways, including with vendor-neutral marketplaces or by the gradual consolidation of providers of AI-based solutions.

    With the expanding use of AI, the issue of trust between AI developers, healthcare professionals, regulators, and patients will become more relevant. It is therefore likely that efforts will intensify to take steps towards strengthening that trust. This will potentially include raising the expected standards of evidence for AI- based solutions (Aggarwal, Sounderajah, et al., 2021; X. Liu et al., 2019; van Leeuwen et al., 2021; Yusuf et al., 2020), making them more transparent through the use and improvement of interpretable AI techniques (Holzinger et al., 2017; Reyes et al., 2020; “Towards Trustable Machine Learning,” 2018), and enhancing techniques for maintaining patient data privacy (G. Kaissis et al., 2021; G. A. Kaissis et al., 2020).

    Furthermore, while most existing regulations stipulate that AI-based algorithms cannot be modified after regulatory approval, this is likely to change in the future. The potential for these algorithms to learn from data acquired after approval and adapt to changing circumstances is a major advantage of AI. Still, frameworks for doing so have thus far been lacking in the healthcare sector. However, promising ideas have recently emerged, including adapting existing hospital quality assurance and improvement frameworks to monitor AI-based algorithms’ performance and the data they are trained on and update the algorithms accordingly (Feng et al., 2022). This will likely require the development of multidisciplinary teams within hospitals consisting of clinicians, IT professionals, and biostatisticians who closely collaborate with model developers and regulators (Feng et al., 2022).

    While the obstacles discussed in previous sections might slow down the adoption of AI in radiology somewhat, the fear of AI potentially replacing radiologists is unlikely to be one of them. A recent survey from Europe showed that most radiologists did not perceive a reduction in their clinical workload after adopting AI-based solutions (European Society of Radiology (ESR), 2022), likely because, at the same time, demand for radiologists’ services has been continuously rising. Studies from around the world have shown that radiology professionals, particularly those with AI exposure and experience, are generally optimistic about the role of AI in their practice (Y. Chen et al., 2021; Huisman et al., 2021; Ooi et al., 2021; Santomartino & Yi, 2022; Scott et al., 2021).

    AI has shown promise in positively impacting virtually every facet of a radiology department’s work - from scheduling and protocolling patient scans to interpreting images and reaching diagnoses. Promising research on AI-based tools in radiology has not yet been widely translated to adoption in routine practice, however, because of a number of complex, partially intertwined issues. Potential solutions exist for many of these challenges, but many of these solutions require further refinement and testing. In the meantime, guidelines are emerging to help potential users of AI-based solutions in radiology navigate the increasing number of commercial products. This encourages their adoption in real-world scenarios, thus allowing their true potential to be uncovered, as well as their weaknesses to be identified and addressed in a safe and effective way. As these incremental improvements are made, these tools will likely evolve to handle more varied data, become integrated into consolidated workflows, become more transparent, and ultimately more useful for increasing efficiency and improving patient care.

    AAMC Report Reinforces Mounting Physician Shortage. (2021). AAMC. https://www.aamc.org/news-insights/press- releases/aamc-report-reinforces-mounting-physician-shortage

    A buyer’s guide to AI in health and care. (2020). NHS Transformation Directorate. https://www.nhsx.nhs.uk/ai-lab/ explore-all-resources/adopt-ai/a-buyers-guide-to-ai-in-health- and-care/

    Advanced AI solutions for radiology. (n.d.). Calantic Website. Retrieved July 3, 2022, from https://aivisions.calantic.com/

    Aggarwal, R., Farag, S., Martin, G., Ashrafian, H., & Darzi, A. (2021). Patient Perceptions on Data Sharing and Applying Artificial Intelligence to Health Care Data: Cross-sectional Survey. Journal of Medical Internet Research, 23(8), e26162. https://doi.org/10.2196/26162

    Aggarwal, R., Sounderajah, V., Martin, G., Ting, D. S. W., Karthikesalingam, A., King, D., Ashrafian, H., & Darzi, A. (2021). Diagnostic accuracy of deep learning in medical imaging: a systematic review and meta-analysis. NPJ Digital Medicine, 4(1), 65. https://doi.org/10.1038/s41746-021-00438-z

    Agnes, S. A., Anitha, J., Pandian, S. I. A., & Peter, J. D. (2019). Classification of Mammogram Images Using Multiscale all Convolutional Neural Network (MA-CNN). Journal of Medical Systems, 44(1), 30. https://doi.org/10.1007/s10916-019-1494-z

    AIF360: A comprehensive set of fairness metrics for datasets and machine learning models, explanations for these metrics, and algorithms to mitigate bias in datasets and models. (n.d.). Github. Retrieved June 11, 2022, from https:// github.com/Trusted-AI/AIF360

    AI for radiology. (n.d.). Retrieved June 26, 2022, from https://grand-challenge.org/
    aiforradiology/?subspeciality=Abdomen&modality=All&ce_ under=All&ce_class=All&fda_class=All&sort_by=last %20 modified&search=

    Akagi, M., Nakamura, Y., Higaki, T., Narita, K., Honda, Y., Zhou, J., Yu, Z., Akino, N., & Awai, K. (2019). Deep learning reconstruction improves image quality of abdominal ultra- high-resolution CT. European Radiology, 29(11), 6163–6171. https://doi.org/10.1007/s00330-019-06170-3

    Akkus, Z., Cai, J., Boonrod, A., Zeinoddini, A., Weston, A. D., Philbrick, K. A., & Erickson, B. J. (2019). A Survey of Deep- Learning Applications in Ultrasound: Artificial Intelligence- Powered Ultrasound for Improving Clinical Workflow. Journal of the American College of Radiology: JACR, 16(9 Pt B), 1318–1328. https://doi.org/10.1016/j.jacr.2019.06.004

    Al-Antari, M. A., Al-Masni, M. A., & Kim, T.-S. (2020). Deep Learning Computer-Aided Diagnosis for Breast Lesion in Digital Mammogram. Advances in Experimental Medicine and Biology, 1213, 59–72. https://doi.org/10.1007/978-3-030-33128-3_4

    Alexander, A., Jiang, A., Ferreira, C., & Zurkiya, D. (2020). An Intelligent Future for Medical Imaging: A Market Outlook on Artificial Intelligence for Medical Imaging. Journal of the American College of Radiology: JACR, 17(1 Pt B), 165–170. https:// doi.org/10.1016/j.jacr.2019.07.019

    Allen, B., Agarwal, S., Coombs, L., Wald, C., & Dreyer, K. (2021). 2020 ACR Data Science Institute Artificial Intelligence Journal of the American College of Radiology: JACR

    Alzubaidi, L., Zhang, J., Humaidi, A. J., Al-Dujaili, A., Duan, Y., Al-Shamma, O., Santamaría, J., Fadhel, M. A., Al-Amidie, M., & Farhan, L. (2021). Review of deep learning: concepts, CNN architectures, challenges, applications, future directions.

    Arbabshirani, M. R., Fornwalt, B. K., Mongelluzzo, G. J., Suever, J. D., Geise, B. D., Patel, A. A., & Moore, G. J. (2018). Advanced machine learning in action: identification of intracranial hemorrhage on computed tomography scans of the head with clinical workflow integration. NPJ Digital Medicine, 1, 9. https://doi.org/10.1038/s41746-017-0015-z

    Ardila, D., Kiraly, A. P., Bharadwaj, S., Choi, B., Reicher, J. J., Peng, L., Tse, D., Etemadi, M., Ye, W., Corrado, G., Naidich, D. P., & Shetty, S. (2019). End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nature Medicine, 25(6), 954–961. https://doi. org/10.1038/s41591-019-0447-x

    Arefan, D., Mohamed, A. A., Berg, W. A., Zuley, M. L., Sumkin, J. H., & Wu, S. (2020). Deep learning modeling using normal mammograms for predicting breast cancer risk. Medical Physics, 47(1), 110–118. https://doi.org/10.1002/mp.13886

    Assess-AI. (n.d.). Retrieved July 2, 2022, from https://www. acrdsi.org/DSI-Services/Assess-AI

    Azer, S. A. (2019). Deep learning with convolutional neural networks for identification of liver masses and hepatocellular carcinoma: A systematic review. World Journal of Gastrointestinal Oncology, 11(12), 1218–1230. https://doi.org/10.4251/wjgo.v11. i12.1218

    Bacchi, S., Zerner, T., Oakden-Rayner, L., Kleinig, T., Patel, S., & Jannes, J. (2020). Deep Learning in the Prediction of Ischaemic Stroke Thrombolysis Functional Outcomes: A Pilot Study. Academic Radiology, 27(2), e19–e23. https://doi. org/10.1016/j.acra.2019.03.015

    Baccianella, S., & Gough, T. (n.d.). Why cloud computing is the best option for hospitals adopting AI. Retrieved June 11, 2022, from https://www.aidence.com/articles/cloud-best-option- imaging-ai/

    Balint, B. J., Steenburg, S. D., Lin, H., Shen, C., Steele, J. L., & Gunderman, R. B. (2014). Do telephone call interruptions have an impact on radiology resident diagnostic accuracy? Academic Radiology, 21(12), 1623–1628. https://doi.org/10.1016/j. acra.2014.08.001

    Baltruschat, I., Steinmeister, L., Nickisch, H., Saalbach, A., Grass, M., Adam, G., Knopp, T., & Ittrich, H. (2021). Smart chest X-ray worklist prioritization using artificial intelligence: a clinical workflow simulation. European Radiology, 31(6), 3837– 3845. https://doi.org/10.1007/s00330-020-07480-7

    Bash, S., Johnson, B., Gibbs, W., Zhang, T., Shankaranarayanan, A., & Tanenbaum, L. N. (2021). Deep Learning Image Processing Enables 40 % Faster Spinal MR Scans Which Match or Exceed Quality of Standard of Care : A Prospective Multicenter Multireader Study. Clinical Neuroradiology. https://doi.org/10.1007/s00062-021-01121-2

    Bash, S., Wang, L., Airriess, C., Zaharchuk, G., Gong, E., Shankaranarayanan, A., & Tanenbaum, L. N. (2021). Deep Learning Enables 60 % Accelerated Volumetric Brain MRI While Preserving Quantitative Performance: A Prospective, Multicenter, Multireader Trial. AJNR. American Journal of Neuroradiology, 42(12), 2130–2137. https://doi.org/10.3174/ajnr.A7358

    Boas, F. E., & Fleischmann, D. (2012). CT artifacts: causes and reduction techniques. Imaging in Medicine, 4(2), 229–240. https://doi.org/10.2217/iim.12.13

    Boland, G. W., Duszak, R., Jr, & Kalra, M. (2014). Protocol design and optimization. Journal of the American College of Radiology: JACR, 11(5), 440–441. https://doi.org/10.1016/j. jacr.2014.01.021

    Bonavita, I., Rafael-Palou, X., Ceresa, M., Piella, G., Ribas, V., & González Ballester, M. A. (2020). Integration of convolutional neural networks for pulmonary nodule malignancy assessment in a lung cancer classification pipeline. Computer Methods and Programs in Biomedicine, 185, 105172. https://doi.org/10.1016/j.cmpb.2019.105172

    Bozkurt, S., Cahan, E. M., Seneviratne, M. G., Sun, R., Lossio- Ventura, J. A., Ioannidis, J. P. A., & Hernandez-Boussard, T. (2020). Reporting of demographic data and representativeness in machine learning models using electronic health records.
    Journal of the American Medical Informatics Association: JAMIA, 27(12), 1878–1884. https://doi.org/10.1093/jamia/ocaa164

    Brown, A. D., & Marotta, T. R. (2018). Using machine learning for sequence-level automated MRI protocol selection in neuroradiology. Journal of the American Medical Informatics Association: JAMIA, 25(5), 568–571. https://doi.org/10.1093/ jamia/ocx125

    Bruls, R. J. M., & Kwee, R. M. (2020). Workload for radiologists during on-call hours: dramatic increase in the past 15 years.
    Insights into Imaging, 11(1), 121. https://doi.org/10.1186/ s13244-020-00925-z

    Buruk, B., Ekmekci, P. E., & Arda, B. (2020). A critical perspective on guidelines for responsible and trustworthy artificial intelligence. Medicine, Health Care, and Philosophy, 23(3), 387–399. https://doi.org/10.1007/s11019-020-09948-1

    Cadario, R., Longoni, C., & Morewedge, C. K. (2021). Understanding, explaining, and utilizing medical artificial intelligence. Nature Human Behaviour, 5(12), 1636–1642. https://doi.org/10.1038/s41562-021-01146-0

    Center for Devices, & Radiological Health. (n.d.). Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices. U.S. Food and Drug Administration; FDA. Retrieved July 2, 2022, from https://www.fda.gov/medical-devices/software- medical-device-samd/artificial-intelligence-and-machine- learning-aiml-enabled-medical-devices

    Char, D. S., Shah, N. H., & Magnus, D. (2018). Implementing Machine Learning in Health Care - Addressing Ethical Challenges. The New England Journal of Medicine, 378(11), 981–983. https://doi.org/10.1056/NEJMp1714229

    Chen, H., Zhang, Y., Kalra, M. K., Lin, F., Chen, Y., Liao, P., Zhou, J., & Wang, G. (2017). Low-Dose CT With a Residual Encoder-Decoder Convolutional Neural Network. IEEE Transactions on Medical Imaging, 36(12), 2524–2535. https://doi.org/10.1109/TMI.2017.2715284

    Chen, Y., Stavropoulou, C., Narasinkan, R., Baker, A., & Scarbrough, H. (2021). Professionals’ responses to the introduction of AI innovations in radiology and their implications for future adoption: a qualitative study. BMC Health Services Research, 21(1), 813. https://doi.org/10.1186/ s12913-021-06861-y

    Choe, J., Lee, S. M., Do, K.-H., Lee, G., Lee, J.-G., Lee, S. M., & Seo, J. B. (2019). Deep Learning-based Image Conversion of CT Reconstruction Kernels Improves Radiomics Reproducibility for Pulmonary Nodules or Masses. Radiology, 292(2), 365–373. https://doi.org/10.1148/radiol.2019181960

    Choi, K. S., Choi, S. H., & Jeong, B. (2019). Prediction of IDH genotype in gliomas with dynamic susceptibility contrast perfusion MR imaging using an explainable recurrent neural network. Neuro-Oncology, 21(9), 1197–1209. https://doi.org/10.1093/neuonc/noz095

    Chong, L. R., Tsai, K. T., Lee, L. L., Foo, S. G., & Chang, P. C. (2020). Artificial Intelligence Predictive Analytics in the Management of Outpatient MRI Appointment No-Shows. AJR. American Journal of Roentgenology, 215(5), 1155–1162. https://doi.org/10.2214/AJR.19.22594

    Cikes, M., Sanchez-Martinez, S., Claggett, B., Duchateau, N., Piella, G., Butakoff, C., Pouleur, A. C., Knappe, D., Biering- Sørensen, T., Kutyifa, V., Moss, A., Stein, K., Solomon, S. D., & Bijnens, B. (2019). Machine learning-based phenogrouping in heart failure to identify responders to cardiac resynchronization therapy. European Journal of Heart Failure, 21(1), 74–85. https://doi.org/10.1002/ejhf.1333

    Ciompi, F., Chung, K., van Riel, S. J., Setio, A. A. A., Gerke, P. K., Jacobs, C., Scholten, E. T., Schaefer-Prokop, C., Wille,
    M. M. W., Marchianò, A., Pastorino, U., Prokop, M., & van Ginneken, B.
    (2017). Towards automatic pulmonary nodule management in lung cancer screening with deep learning. Scientific Reports, 7, 46479. https://doi.org/10.1038/srep46479

    Clinical radiology UK workforce census 2019 report. (2019). https://www.rcr.ac.uk/publication/clinical-radiology-uk- workforce-census-2019-report

    Cloud security for healthcare services. (2021, January 14). ENISA. https://www.enisa.europa.eu/publications/cloud- security-for-healthcare-services/

    CONSORT-AI and SPIRIT-AI Steering Group. (2019). Reporting guidelines for clinical trials evaluating artificial intelligence interventions are needed. Nature Medicine, 25(10), 1467–1468. https://doi.org/10.1038/s41591-019-0603-3

    Curated marketplace. (2018, May 22). Blackford. https://www.blackfordanalysis.com/applications/

    Dance, A. (2021). AI spots cell structures that humans can’t. Nature. 592 (7852), 154–155.

    Dantas, L. F., Fleck, J. L., Cyrino Oliveira, F. L., & Hamacher, S. (2018). No-shows in appointment scheduling - a systematic literature review. Health Policy, 122(4), 412–421. https://doi.org/10.1016/j.healthpol.2018.02.002

    Deák, Z., Grimm, J. M., Treitl, M., Geyer, L. L., Linsenmaier, U., Körner, M., Reiser, M. F., & Wirth, S. (2013). Filtered back projection, adaptive statistical iterative reconstruction, and a model-based iterative reconstruction in abdominal CT: an experimental clinical study. Radiology, 266(1), 197–206. https://doi.org/10.1148/radiol.12112707

    Dembrower, K., Liu, Y., Azizpour, H., Eklund, M., Smith, K., Lindholm, P., & Strand, F. (2020). Comparison of a Deep
    Learning Risk Score and Standard Mammographic Density Score for Breast Cancer Risk Prediction. Radiology, 294(2), 265–272. https://doi.org/10.1148/radiol.2019190872

    Do, B. H., Langlotz, C., & Beaulieu, C. F. (2017). Bone Tumor Diagnosis Using a Naïve Bayesian Model of Demographic and Radiographic Features. Journal of Digital Imaging, 30(5), 640–647. https://doi.org/10.1007/s10278-017-0001-7

    Dou, Q., Yu, L., Chen, H., Jin, Y., Yang, X., Qin, J., & Heng, P.-A. (2017). 3D deeply supervised network for automated segmentation of volumetric medical images. Medical Image Analysis, 41, 40–54. https://doi.org/10.1016/j. media.2017.05.001

    Eche, T., Schwartz, L. H., Mokrane, F.-Z., & Dercle, L. (2021). Toward Generalizability in the Deployment of Artificial Intelligence in Radiology: Role of Computation Stress Testing to Overcome Underspecification. Radiology. Artificial Intelligence, 3(6), e210097. https://doi.org/10.1148/ryai.2021210097

    England, N. H. S., & Improvement, N. H. S. (2019). NHS diagnostic waiting times and activity data. NHS. https://www. england.nhs.uk/statistics/wp-content/uploads/sites/2/2021/12/ DWTA-Report-October-2021_M43D4.pdf

    Esmaeilzadeh, P. (2020). Use of AI-based tools for healthcare purposes: a survey study from consumers’ perspectives. BMC Medical Informatics and Decision Making, 20(1), 170. https://doi. org/10.1186/s12911-020-01191-1

    Esses, S. J., Lu, X., Zhao, T., Shanbhogue, K., Dane, B., Bruno, M., & Chandarana, H. (2018). Automated image quality evaluation of T2 -weighted liver MRI utilizing deep learning architecture. Journal of Magnetic Resonance Imaging: JMRI, 47(3), 723–728. https://doi.org/10.1002/jmri.25779

    European Society of Radiology (ESR). (2022). Current practical experience with artificial intelligence in clinical radiology:
    a survey of the European Society of Radiology. Insights into Imaging, 13(1), 107. https://doi.org/10.1186/s13244-022- 01247-y

    Ferlay, J., Colombet, M., Soerjomataram, I., Dyba, T., Randi,G., Bettio, M., Gavin, A., Visser, O., & Bray, F.(2018). Cancer incidence and mortality patterns in Europe: Estimates for 40 countries and 25 major cancers in 2018. European Journal of Cancer, 103, 356–387. https://doi.org/10.1016/j.ejca.2018.07.005