AI-Driven Technology Enhances Accuracy and Speed in Monitoring Heart Stent Healing

A groundbreaking advancement in cardiology includes the development of DeepNeo, an artificial intelligence (AI) algorithm that automates the assessment of coronary stent healing via optical coherence tomography (OCT) imaging. Developed through a collaboration between Helmholtz Munich, the Technical University of Munich, and the TUM University Hospital, DeepNeo promises to transform post-stent implantation monitoring.

Each year, over 3 million individuals worldwide receive stents to remedy blocked coronary arteries caused by heart disease. However, evaluating how well the tissue heals after stent placement remains a complex and time-consuming process. Irregular healing, such as excessive tissue growth or deposits forming over the stent, can lead to critical issues like re-narrowing or complete vessel occlusion. Traditional analysis of OCT images for these patterns involves manual review, which is impractical for routine clinical use.

DeepNeo addresses this challenge by providing an automated, highly precise assessment of stent healing, matching the accuracy of experienced clinicians but in a fraction of the time. The AI model classifies tissue types with an accuracy comparable to experts, identifying areas like homogeneous neointima and regions affected by neoatherosclerosis, which are crucial for patient management. It also delivers exact measurements of tissue thickness and stent coverage, offering valuable insights for personalized treatment.

The research team trained DeepNeo using 1,148 OCT images from 92 patients, meticulously annotated to distinguish various tissue growth patterns. Validation in animal models demonstrated that DeepNeo correctly identified unhealthy tissue in 87% of cases, aligning closely with laboratory gold standards. When tested on human scans, the AI maintained high accuracy, supporting its potential for clinical adoption.

"With DeepNeo, we can achieve automated, standardized, and reliable analysis of stent and vascular healing, previously only possible through extensive manual effort," said Valentin Koch, first author of the study. Experts believe that integrating AI tools like DeepNeo into clinical workflows can support quicker, more informed decisions, ultimately improving cardiovascular outcomes.

The project has secured a Helmholtz Innovation Grant, and patent applications have been filed. Industry partnerships are underway to facilitate the transition of DeepNeo from research to healthcare practice. Cardiologists from TUM emphasize that such AI systems could significantly reduce healthcare costs while enabling more personalized and effective treatments, paving the way for smarter, faster cardiovascular care.