AI-assisted adhesion detection on cine-MRI

Chronic abdominal pain after surgery has a significant impact on patients and healthcare. Up to 20% of patients develop postoperative chronic pain and in half of these cases, the symptoms may be caused by adhesions. This results in approximately 15,000 patients affected by painful adhesions each year in The Netherlands alone. Surgery is the only option to remove adhesions, but it is often difficult to judge who would benefit from it. Cine-MRI is a relatively novel modality, which holds great potential for non-invasively creating assessing the presence and location of intra-abdominal adhesions. This allows for better selection of patients suitable for surgery and better surgery planning, resulting in less laparoscopies, done better.

In a recent observer study we found that there is ample room for improvement in both diagnostic performance and consistency, especially for observers with relative little experience. We have developed an Artificial Intelligence (AI) models that can sensitively (~80%) detect and locate adhesions on cine-MRI. This would tremendously help in finding nearly all potential adhesion locations, but still requires discarding false positives. In this reader study, we aim to validate AI assistance for adhesion detection on cine-MRI. The outcome of this study will be published and a positive outcome would help bring cine-MRI to more clinics worldwide.

Joining the study

We are looking for observers! Specific experience with cine-MRI or adhesion detection is not necessary.

Partcipation requires reading 50 mid-sagittal cine-MRI slices. The total estimated time expenditure is estimated to be 4 hours, which includes a small demo study. The study runs fully online in a browser of your choice in the reader study environment of grand-challenge and can be completed at your own pace in April and May. See the video below for an example of the annotation workflow. Joining the study as a reader will result in a (consortium) co-authorship on the final publication.

Interested in joining the study, or want to know more? Contact bram.dewilde@radboudumc.nl

The animation shows the reader study workflow. Initially, the cine-MRI slice is assessed using imaging data only. Annotations are made using a yellow brush tool. Then, an AI heatmap is shown (purple) to help spot adhesions. This allows to study the effect of AI assistance on cine-MRI for adhesion detection.