Automated OCT Image Analysis for Coronary Artery Disease
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing intracoronary optical coherence tomography (OCT) images are time-consuming, prone to variability, and limited in predicting high-risk plaques and disease progression, with suboptimal accuracy and lack of validation against clinical trial data.
Innovation Solution
An automated computer-implemented method using a sequence of neural networks for OCT image analysis, including classification, segmentation, artifact correction, and feature measurement, to distinguish healthy and diseased tissue, correct artifacts, and assess plaque composition, with predictive models for therapy efficacy and patient outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual frame selection and measurement in specialized core laboratories is used, then detailed OCT analysis can be performed, but the process is time-consuming and requires offline processing
Solution Approach 1:
The patent replaces manual mechanical analysis processes with an automated computer-implemented system using machine learning algorithms. The system automatically performs frame selection, measurement, and analysis of OCT images without requiring offline manual processing in core laboratories, thereby maintaining measurement precision while dramatically reducing analysis time.
Solution Approach 2:
The automated system enables self-service analysis by processing OCT images independently without requiring specialized core laboratory personnel. The machine learning model autonomously performs segmentation, artifact detection, and measurement tasks that previously required expert manual intervention, eliminating the time loss associated with manual processing while maintaining analytical accuracy.
2Productivity
If automated OCT analysis systems are used, then processing time is reduced, but accuracy in identifying disease progression and predicting patient outcomes is insufficient
Solution Approach 1:
The patent incorporates feedback mechanisms where the automated system's predictions and measurements are continuously refined based on comparison with ground truth data from clinical trials and core laboratories. The machine learning model learns from feedback to improve its accuracy in identifying disease progression and predicting patient outcomes while maintaining high processing speeds.
Solution Approach 2:
The system performs preliminary automated analysis to identify potential disease features and patterns before final validation. By pre-processing and flagging suspicious areas using machine learning, the system prepares data for more focused accuracy-critical analysis, thereby achieving both high productivity and measurement precision in disease prediction.
3Productivity
If models are trained on small or highly selected training datasets, then model development is faster, but generalizability and accuracy in real-world clinical practice are limited
Solution Approach 1:
The patent performs preliminary actions by curating and preparing large, diverse training datasets from multiple clinical trials and real-world clinical practice before model development. This upfront preparation of comprehensive data ensures that models are trained on representative samples, improving generalizability while maintaining efficient model development through automated data processing pipelines.
Solution Approach 2:
The patent creates universal training datasets that encompass diverse patient populations, disease stages, and imaging conditions from multiple sources. By training models on this universal data that reflects real-world variability, the system achieves both rapid model development and high generalizability across different clinical settings and patient groups.
Data Source
AI summary
Disclosed is a method for analyzing a set of images of a coronary artery tissue. The method comprises segmenting the images for the presence of normal artery features and those associated with OCT, correcting artifacts, and optimizing the images. The method further comprises segmenting the diseased tissue into distinct tissue types, and measuring features of interests of the segmented tissue types. The method further comprises compiling a first set of measurements for each identified feature of interest at a first time, and a second set of measurements at a second time subsequent to the first time. The method further comprises determining changes in the coronary artery tissue, indicative of progression or regression of a diseased state, or prediction of multiple adverse cardiovascular events (MACE) such as cardiac death or myocardial infarction.