Time to Integrate AI into MRI? Debate Continues

These days it seems like the pace of living is continually ramping up. As one blogger expresses, “We live in a culture of hurry, which assumes that doing something more quickly is always better than doing it more slowly. At the root of that is a veneration of productivity: the more quickly you can do something, the more you can get done, and the more you get done, presumably the more productive you are.”

When it comes to interpreting MRI scans of the prostate, how productive a radiologist is matters to both the referring doctor and the patient. The turnaround time matters, because if prostate cancer (PCa) is present it doesn’t wait for a diagnosis to continue the activity of the tumor. Whether it’s slow-growing or aggressive, the imaging appointment doesn’t come with a cancer “pause” button.

Meanwhile, the reviewing and diagnostic tasks can’t be done at a finger snap. One of our radiology colleagues, Dr. Maarten de Rooij at Radboud University (Nijmegen, the Netherlands), points out that the process “… is time-consuming, involving the generation of detailed reports, contouring lesions for targeted biopsies or (focal) treatments, and comparing current scans with previous ones for active surveillance, follow-up, and disease recurrence.”^[i] Now that multiparametric MRI (mpMRI) is widely incorporated in the PCa detection/diagnostic pathway, radiologist readers are increasingly inundated with images to review and report on.

It’s no wonder, then, that hopes are high over the potential for Artificial Intelligence (AI) to come to their rescue. Our blog posts include descriptions of recent studies demonstrating the efficiency and accuracy of new AI software as these models are developed. Still, de Rooij and others encourage putting the brakes on hopes and enthusiasm. Their concerns include such points as

• Current models tend to narrowly focus on identifying only suspicious areas
• AI systems don’t generate reports on incidental findings or benign conditions that can cause abnormally high PSA results
• In real life, AI does not live up to promised stress relief for overloaded radiologists
• It’s not yet clear who pays for expensive AI software plus the time-consuming learning curve of radiologists who will use it – Insurance? Healthcare providers? Patients?

On the other hand, supporters give prostate centers a green light to pilot programs using currently approved models that are on the market. For example, Dr. Tobias Penzkofer (Department of Radiology, Charité Universitätsmedizin Berlin) believes the time is now to implement AI systems to assist MRI readers. Especially as a preliminary read, he has pointed out “convenience factors … such as automatic prostate segmentation and volumetry, which allows for automatic calculation of PSA density, potentially with higher fidelity compared with the traditional ellipsoid formula. Also, automatic structured report creation from AI annotations and segmentation, even if revised by the reader, can contribute to the clarity and consistency of reports …”^[ii] Clearly, he champions jumping into integration of AI tools that can assist detection while leaving final judgment to experienced humans who can correlate additional case information to make bigger-picture determinations.

While our Center understands the considerations raised by those who urge caution, we believe that responsible use of existing and future AI tools can only benefit clinicians and the PCa patients we serve.

NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.

[i] Philip Ward. “Debate intensifies over AI use for prostate MRI.” AuntMinnieEurope, Jul. 11, 2024. https://www.auntminnieeurope.com/clinical-news/mri/article/15679376/debate-intensifies-over-ai-use-for-prostate-mri
[ii] Ibid.