We’re thrilled to share our latest publication from the CARA Lab: the PECTUS-AI study. Thin-cap fibroatheroma (TCFA) is considered a high-risk feature linked to adverse outcomes in patients with coronary artery disease. Assessing TCFA on intracoronary OCT is critical but notoriously difficult: it requires expert readers, takes hours, and is inconsistent across observers. In the PECTUS-AI study, we tested whether our deep learning algorithm, OCT-AID, could enhance this assessment. Trained for full-vessel segmentation and plaque quantification, OCT-AID analyzes every frame of an OCT pullback, measuring fibrous cap thickness, detecting lipid cores, and automatically evaluating TCFA.

Applied to OCT images from 414 myocardial infarction patients, AI detected TCFA in 34.5% of cases (vs. 30.0% by core lab experts). More importantly, AI-based TCFA was significantly associated with adverse outcomes at two years, with risk nearly doubling when TCFA was present in target lesions. When OCT-AID analyzed the entire imaged segment, prognostic power was even stronger (HR 5.5), with a negative predictive value ~98%.

Unlike manual review, which is slow and variable, AI delivers standardized, reproducible, and comprehensive plaque assessment within minutes. Our findings show that AI doesn’t just match expert interpretation, it provides superior risk prediction by uncovering vulnerability throughout the coronary artery. This brings us closer to real-time, AI-assisted decision-making in the cath lab, where high-risk plaques can be recognized and treated before they cause harm. The impact of this research was widely recognized, with coverage in national media, including RTL Nieuws, RTL Tonight, De Gelderlander, and BNR Nieuwsradio.

Read the full article here!

Image source: Volleberg, Rick HJA, et al. “Artificial intelligence-based identification of thin-cap fibroatheromas and clinical outcomes: the PECTUS-AI study.” European Heart Journal (2025): ehaf595.