We’re excited to share a new study from the CARA Lab that will be presented as a poster at ISBI 2026.

In this work, led by Leah Heil and Simone Saitta, we build on our OCT-AID framework for intracoronary OCT imaging. OCT-AID can automatically analyze plaque in the coronary arteries, but until now it did not indicate how sure it was about its own results. In the new study, we add a way for the AI to show its confidence across the image, highlighting regions where its assessment may be less reliable.

We show that areas flagged as uncertain by the AI are indeed more difficult. They are where the analysis is less accurate and where human experts are more likely to disagree as well. This turns uncertainty into a practical tool: it can help focus human review on the most challenging regions and support safer, more transparent use of AI in intracoronary OCT.

We look forward to presenting this work at ISBI 2026 on behalf of the CARA Lab!