Coronary CT Artery Wall Segmentation for CAD Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying coronary artery disease using predicted radius from an artery centreline provide limited information and are not effective in characterizing the disease.
Innovation Solution
A method and system using machine learning to analyze contrast cardiac CT data, identify centreline seed points, and determine inner and outer artery walls to detect and characterize coronary artery disease, including stenosis and high-risk plaques, by employing techniques such as radiodensity testing, branch detection, and vessel wall segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If predicted radius from artery centreline is used to identify coronary artery disease, then disease identification can be performed, but limited information is provided in relation to disease characterization
Solution Approach 1:
The patent segments the coronary artery into multiple components: centreline extraction, inner wall detection, outer wall detection, and lumen area calculation. This segmentation allows each component to be analyzed independently and provides comprehensive disease information including stenosis percentage and plaque characterization, resolving the contradiction between information completeness and system complexity.
Solution Approach 2:
The patent transitions from one-dimensional centreline radius measurement to three-dimensional vessel wall segmentation. By detecting inner and outer walls in 3D space and calculating cross-sectional lumen areas, the system provides multi-dimensional disease characterization including volumetric stenosis assessment and plaque morphology, significantly increasing information content.
2Measurement precision
If machine learning is used to identify centreline seed points and vessel walls, then disease characterization accuracy is improved, but processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by first identifying centreline seed points and establishing vessel centrelines before detecting inner and outer walls. This sequential preliminary processing guides subsequent wall detection algorithms, improving accuracy while reducing overall processing time compared to simultaneous analysis of all components.
Solution Approach 2:
The patent uses machine learning models as intermediary components that process CT data through multiple stages: centreline detection model, inner wall detection model, and outer wall detection model. Each model acts as an intermediary processor, transforming raw CT data into structured vessel geometry information, thereby improving precision through specialized processing while managing computational load through modular architecture.
3Loss of information
If multi-stage machine learning analysis is performed to detect inner and outer walls, then disease characterization is enhanced, but system complexity increases
Solution Approach 1:
The patent divides the complex vessel wall detection problem into separate stages: centreline identification, inner wall detection, and outer wall detection. Each stage uses specialized machine learning models trained for specific tasks, allowing independent optimization and simplifying the overall system architecture while comprehensively capturing vessel wall information.
Solution Approach 2:
The patent employs a unified machine learning framework that handles multiple functions through a consistent processing pipeline. The same architectural principles and training methodologies are applied across different detection stages, creating a universal system that can accurately detect various vessel wall features while managing complexity through standardized procedures.
Data Source
AI summary
A method and corresponding system of identifying coronary artery disease. The method comprises receiving contrast cardiac CT data indicative of a contrast cardiac CT scan carried out on a patient, and analysing the contrast cardiac CT data using machine learning to identify a plurality of seed points in the contrast cardiac CT data expected to correspond to locations in cardiac arteries of the patient. The method also comprises producing data indicative of transverse image slices of the cardiac arteries of the patient using the contrast cardiac CT data and the identified seed points, analysing the transverse image slice data using machine learning to produce inner artery wall data and outer artery wall data indicative of predicted respective inner and outer walls of the coronary arteries of the patient, and identifying presence of coronary artery disease using the predicted inner and/or outer walls of the coronary arteries of the patient.