New Technology Accelerates and Improves the Accuracy of Coronary Stent Healing Analysis

Proper Healing Is Key

More than three million patients worldwide undergo coronary stent implantation each year due to ischemic heart disease. While the procedure restores blood flow in narrowed arteries, the success of treatment also depends on proper healing—specifically, the growth of new neointimal tissue.

If this process is disrupted—such as by excessive hyperplasia or calcification—there is an increased risk of restenosis or even complete vessel occlusion.

Accuracy, Speed, and More Insights

OCT is a precise imaging method that allows detailed visualization of the vessel wall and the stent. However, its use in clinical practice is limited by the time-consuming nature of manual analysis. Interpreting hundreds of images can take hours, even for experienced specialists. This is where AI comes into play.

A research team from Helmholtz Zentrum München and the University Hospital of the Technical University of Munich developed the DeepNeo algorithm, which enables automated stent healing analysis based on OCT images.

DeepNeo not only identifies various types of healing patterns with accuracy comparable to human experts but also does so within seconds. It additionally provides quantitative data, such as neointimal thickness or stent coverage, which can aid in further treatment decisions. 

“We can now achieve automated, standardized, and highly accurate analysis of stent and vessel wall healing—something that previously required extensive manual work. DeepNeo is as accurate as a physician but much faster,” explains Dr. Valentin Koch, the study’s first author.

Tailored Treatment for Every Patient

The algorithm was trained on 1,148 OCT images from 92 examinations in patients. All images were manually annotated to classify various forms of neointimal growth. DeepNeo was then tested on an animal model, where it correctly identified pathological tissue in 87% of cases compared to the laboratory gold standard. It also showed high accuracy when analyzing images from human patients.

This tool could become part of a broader AI-driven care system. According to experts, it has the potential to streamline care, reduce costs, and above all enable personalized treatment—based specifically on how an individual’s vessels are healing.

Editorial Team, Medscope.pro

Sources:

1. Koch V., Holmberg O., Blum E. et al. Deep learning model DeepNeo predicts neointimal tissue characterization using optical coherence tomography. Commun Med 5, 124 (2025), doi: 10.1038/s43856-025-00835-5.

2. AI-Powered Analysis of Stent Healing. Helmholtz Munich, April 24, 2025. Available at: www.helmholtz-munich.de/en/newsroom/news-all/artikel/ai-powered-analysis-of-stent-healing