An artificial intelligence (AI)-based optical coherence tomography (OCT) analysis detected patients at increased risk for adverse cardiovascular outcomes and outperformed core laboratory-based OCT analysis, according to results of the PECTUS-AI study.

The findings were published in the European Heart Journal and presented at the European Society of Cardiology (ESC) Congress 2025 by Rick H.J.A. Volleberg, MD, with Radboud University Medical Center, Nijmegen, the Netherlands.

Jos Thannhauser, PhD, an assistant professor at Radboud University Medical Center, told Medscape Medical News that AI-based OCT could “stimulate broader adoption of OCT and help us fully leverage the potential of this information-rich imaging modality.”

Giulio Guagliumi, MD, interventional cardiologist at IRCCS Galeazzi-Sant'Ambrogio Hospital in Milan, Italy, added that the study “brings exciting new data on how AI may impact the practice and finally convince more interventional cardiologists to use imaging in a proper way.”

AI-Based Thin-Cap Fibroatheroma (TCFA) Tops Core Lab Review

Thannhauser said that he and fellow researchers had previously developed and validated AI algorithms to automatically analyze and interpret OCT data.

“The next logical step was not just to show that the algorithm performs well technically but also to investigate whether its output has clinical relevance,” he said, adding that the research team performed the analysis to answer the question of whether identification of AI-based TCFA relates to patient outcomes.

The PECTUS-AI study, which was a secondary analysis of the prospective observational PECTUS-obs study, included patients with myocardial infarction who underwent OCT of all fractional flow reserve-negative nonculprit, or target, lesions. An independent core laboratory and the OCT-AID segmentation algorithm assessed OCT images for the presence of TCFA.

The researchers defined the primary endpoint as the composite of all-cause death, nonfatal myocardial infarction, or unplanned revascularization at 2 years (plus or minus 30 days), excluding procedural and stent-related events.

Of the 438 patients enrolled in PECTUS-obs, 414 patients (lesion, n = 488) were eligible for the present analysis. The average patient age was 63 years, and 80.9% were men.

The researchers observed AI-TCFA in 34.5% of patients and core laboratory TCFA in 30% of patients. They reported that AI-TCFA within the target lesion was linked to the primary endpoint (hazard ratio [HR], 1.99; 95% CI, 1.02-3.90; P = .04), but identified no significant association between the primary outcome and core laboratory TCFA (HR, 1.67; 95% CI, 0.84-3.30; P = .14).

Moreover, when assessing the complete pullback, the researchers found an even stronger association between AI-TCFA and the primary endpoint (HR, 5.50; 95% CI, 1.94-15.62; P < .001; negative predictive value, 97.6%; 95% CI, 94.0%-99.3%).

‘The Complete Picture’

The greater prognostic impact observed with AI compared with target lesion evaluation only, Thannhauser believes, is the result of the fundamental difference in scope between the two approaches. “Traditional target lesion analysis is based on the angiogram,” he said. “Core labs scroll through the section of the OCT pullback corresponding to the angiographic lesion, simply because analyzing the entire vessel manually would take too much time.”

“However,” he added, “thin-cap fibroatheromas are not restricted to the target lesion — they can occur anywhere along the vessel. With AI, we can automatically analyze the full pullback, vessel-wide, in a reproducible way. In our study, we found that high-risk areas were also present outside the target lesion, and incorporating these additional findings clearly added prognostic value. That’s most likely why the AI-based approach demonstrated stronger predictive power: It captures the complete picture, not just the angiographic hot spot.”

If these preliminary data are confirmed in appropriately designed studies and these AI tools are made available to cath labs, Guagliumi said, “we might have real-time actionable tools to speed, simplify, and make semiautomatic the coronary analysis, personalizing the treatment and the prognostication.”

‘A Pivotal Step’

In an interview with Medscape Medical News, Ziad A. Ali, MD, DPhil, a professor of cardiology at New York Institute of Technology, said that the PECTUS-AI study has two major clinical implications.

“First, automated analysis may overcome interobserver variability and the impracticality of manual frame-by-frame review, enabling standardized and reproducible identification of high-risk plaques in real-world practice,” Ali said. “Second, the finding that AI-based assessment of the entire imaged segment — rather than only target lesions — carries superior prognostic value underscores the systemic nature of plaque vulnerability.

“For operators,” he continued, “this suggests a future where real-time, AI-driven OCT interpretation could guide patient-level risk stratification and potentially trigger preventive interventions. As someone deeply engaged in intravascular imaging, I see this work as a pivotal step toward integrating AI into daily cath lab workflows, bridging precision imaging with clinical decision-making.”

The study was partly funded by Abbott Vascular. Thannhauser, Guagliumi, and Ali reported having no relevant financial relationships.

Brian Ellis is a freelance writer and editor who lives in Southwest Virginia.