Artificial Intelligence spreads into radiology – changing not only the way diseases are diagnosed, but also how management and research are done.
AI already works with applications on many levels in radiology – clinically, scientifically and management-wise. Calantic and Calantic SPARK aim at integrating AI tools into clinically used systems.
Presenters: Adam Kirell, ‘Within Health’, New York, USA; Todd Johnson, VIDA, Minneapolis, USA
Artificial Intelligence (AI) is a driver for various areas: It pushes quantitative imaging, i.e. leverages huge amounts of data still hidden in images. It also aims at automating management processes in radiology like scheduling follow-up visits.
During an open conversation at RSNA 2022, chief officers from two start-ups talked about AI. Both are part of CalanticTM SPARK, Bayer’s marketplace for innovative companies in the medical imaging AI space.
Automated Follow-Up – The within health Approach
Adam Kirell, CEO of ‘Within Health’®, explained how his company facilitates an important part of patient management in radiology: His application tries to make sure that patients get their clinically recommended follow-up.
Currently, approximately 50% of all patients in the US miss out on their follow-up exams. This not only delays diagnosis, but a missed follow-up exam might contribute to litigation risk. Part of the problem are oftentimes manual and fragmented back-office workflows in these institutions.
Kirell’s system automates these processes by:
- Finding patients lost on follow-up by browsing radiological reports with natural language processing
- Communicating with these patients and the care-team
- Directing these patients to the scheduling team
- Following-up on the appointment and
- Making sure the case can be closed after exams have been performed OR escalating the case
This automated process has highly increased the efficiency of output coordinators, i.e. the staff organizing radiology back-offices: “One output coordinator can now handle almost three times as many patients as before”, said Kirell. This result is particularly important, as many institutions suffer from a lack of staff.
In order to generate more benefit for the radiology ecosystem, his application needs to be integrated smoothly with other applications – which is a key factor for him to be part of CalanticTM SPARK. With Calantic SPARKTM, he expects to knock down barriers like the integration with other applications or technical bandwidth.
Leveraging Imaging Data – The VIDA Approach
Todd Johnson, CTO of VIDA®, talked about AI from a more disease-oriented angle - VIDA® processes chest CT images. The company develops AI-derived imaging biomarkers, i.e. markers for lung diseases extracted from different imaging metrics. These biomarkers may support physicians – pulmonologists or radiologists – in order to take a diagnostic or treatment decision.
One of Johnson’s examples was the use of one biomarker as a predictor for the success of an endobronchial lung resection procedure. This biomarker is now already used in the daily clinical practice at the University of Wisconsin. Other biomarkers are already used to predict interstitial lung disease.
The huge amount of collected imaging data includes vast numbers of biomarkers per clinical time point – what Johnson calls an “imaging biobank”. AI now uses this data to look at the progression of different diseases and model diseases.
The data is also used to optimize and accelerate clinical trials: the technology can be used to create e.g. a synthetic control arm, like “digital twin”. In the end, the biobank reduces the number of subjects required for a new clinical trial.
Regarding Calantic® SPARK, Johnson sees two major benefits of this global and curated platform: He believes that with a partner like Bayer, new applications might gain easier access to the global market. In addition, there might be a chance to build relationships with key academic facilities worldwide.
Presentation Title: Collaboration Sparks Innovation: How Bayer is Engaging with the Startup Ecosystem: Bayer Digital Solutions
Source: RSNA 2022
Last update: 1st February 2023