In the first part of a two-part interview, Nina Kottler, M.D. offers insights and perspective on the recently issued guidance from the FDA on AI-enabled devices and how it may impact developers in the radiology field.
The recently issued draft guidance from the Food and Drug Administration (FDA) on submissions for AI-enabled software reflects more of an emphasis on the total product life cycle for AI products, according to Nina Kottler, M.D., the Associate Chief Medical Officer of Clinical AI for Radiology Partners.
Accordingly, in a recent interview, Dr. Kottler noted that a key cornerstone of the FDA guidance is continuous monitoring of AI software, which may be challenging for smaller AI developers in radiology. She suggested that perhaps an independent group could assist with continuous monitoring of AI software.
“ … What if we had a separate group that could not only help out the AI vendors that maybe couldn't afford to do it, or don't have access to all the data, or (are) not mature enough yet to be there?” posited Dr. Kottler, the Associate Chief Medical Officer of Clinical AI for Radiology Partners.
While Dr. Kottler said the FDA guidance’s increased emphasis on transparency of the training and development of AI software is an important priority, the resulting information needs to be conveyed in a practical way for end-user radiologists.
“Now as an end user, as I'm going through my day looking at 100 cases or 200 cases, many of which are using AI, am I going to that document to look and say, ‘Is this patient and is this exam something that (the AI software) was very specifically trained on? No, that's not realistic. So how do you translate these guidelines into something that's more realistic for the end user? That's going to be the important component. That's the part that I really care about,” maintained Dr. Kottler.
(Editor’s note: For related content, see “Current and Emerging Insights on AI in Breast Imaging: An Interview with Mark Traill, MD, Part 3,” “Diagnostic Imaging’s Top Five AI Content of 2024” and “Nina Kottler, MD, MS on Adoption and Impact of AI in Radiology.”)
For more insights from Dr. Kottler, watch the video below.
Could AI-Powered Abbreviated MRI Reinvent Detection for Structural Abnormalities of the Knee?
April 24th 2025Employing deep learning image reconstruction, parallel imaging and multi-slice acceleration in a sub-five-minute 3T knee MRI, researchers noted 100 percent sensitivity and 99 percent specificity for anterior cruciate ligament (ACL) tears.
Meta-Analysis Shows Merits of AI with CTA Detection of Coronary Artery Stenosis and Calcified Plaque
April 16th 2025Artificial intelligence demonstrated higher AUC, sensitivity, and specificity than radiologists for detecting coronary artery stenosis > 50 percent on computed tomography angiography (CTA), according to a new 17-study meta-analysis.