AI-based Method and Tool (DeepNeo) for Intravascular Optical Coherence Tomography Image Analysis of Coronary Artery Neointimal Tissue
Keywords
Intravascular Optical Coherence Tomography, IV-OCT, Percutaneous coronary interventions, PCI, Artificial Intelligence, Deep Learning, Neointima
Invention Novelty
Researchers of Helmholtz Munich and of Deutsches Herzzentrum München have developed an AI-based method that allows analysis of Intravascular Optical Coherence Tomography (IV-OCT) pullback image series of coronary arteries after Percutaneous Coronary Intervention (PCI). The researchers implemented a corresponding deep learning architecture and trained their model with patient data to come up with an AI-based tool (DeepNeo) for IV-OCT pullback image series analysis to classify the neointimal tissue, which is the newborn tissue covering the stent surface inside treated coronary arteries.
DeepNeo has been developed as an academic tool for research purposes and does not fulfil the regulatory requirements of a medical device.
Value Proposition
PCIs are among the most common procedures in cardiovascular medicine. Despite development of long-term risks such as in-stent restenosis, a tool for standardized, automatic analysis of vascular healing is lacking. To the best of our knowledge, commercially available OCT image analysis tools lack the ability to characterize the neointimal tissue and thus miss important features in post-procedure monitoring. The method implemented by DeepNeo aims to fill this gap and the tool characterizes neointimal tissue and provides insights on vessel morphology in a quick and standardized fashion. This has the potential to automate and standardize the current practice for patients undergoing OCT after PCI and may ultimately help to improve and/or to standardize post-PCI decision-making.
Technology Description
DeepNeo, an AI-based tool, relies on two trained deep neural networks that segment features of coronary vessels and that classify neointimal tissue, respectively. DeepNeo can analyze the neointimal tissue composition of an IV-OCT pullback image series (typically consisting of 300 - 400 frames) and visualizes neointima characteristics for each quadrant of each image of the pullback series.
Commercial Opportunity
The technology is available for in-licensing.
Development Status
A proof of principle has been shown.
Patent Situation
A priority establishing patent application and an international patent application have been filed (EP4475072A1, WO2024251852A1).
Further Reading
Koch V, et al., Deep Learning Model DeepNeo predicts neointimal tissue characterization using optical coherence tomography, Communications Medicine 2025, 5:124
Institute of Origin
