Coronary Ostia Detection in Cardiac CT Volumes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting coronary ostia in cardiac CT volumes are not robust, especially under imaging artifacts and severe obstructions, and perform poorly in detecting bypass coronary ostia, often requiring aorta segmentation and user initialization.
Innovation Solution
The method employs marginal space learning (MSL) to detect native coronary ostia without segmenting the aorta, and constrains the search for bypass ostia to the anterior surface of the ascending aorta, using a trained classifier and connected component analysis to achieve high detection rates with low false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional methods segment the aorta before detecting coronary ostia, then detection can be performed, but the method is not robust under imaging artifacts and severe obstruction
Solution Approach 1:
The method performs preliminary localization of the aortic root and coronary ostia using a hierarchical approach: first localizing the aortic root center, then defining a search volume, and finally detecting ostia within this constrained region. This preliminary structuring of the search space makes the detection more robust to artifacts and obstructions by reducing the area requiring detailed segmentation.
Solution Approach 2:
The detection process is segmented into distinct hierarchical stages: (1) aortic root localization, (2) search volume definition, (3) ostia detection within the volume, and (4) validation. This segmentation allows each stage to be optimized independently and makes the overall system more robust by isolating errors to specific stages rather than requiring complete aorta segmentation.
2Adaptability or versatility
If conventional methods are used for coronary ostia detection, then native ostia can be detected, but bypass coronary ostia detection performs poorly
Solution Approach 1:
The detection algorithm is designed to be universal for both native and bypass coronary ostia. It detects all ostia openings within the search volume without requiring pre-specification of ostia type, using the same geometric and intensity-based criteria for both native and bypass ostia, thereby achieving high accuracy for bypass ostia detection.
Solution Approach 2:
The method dynamically adapts the search volume and detection parameters based on the detected aortic root position and patient-specific anatomy. The search volume is dynamically positioned and sized around the aortic root, allowing the algorithm to accommodate variations in ostia location and appearance for both native and bypass grafts.
3Ease of operation
If automatic seeding methods are used to initialize coronary artery centerline extraction, then user specification is eliminated, but detection accuracy decreases under severe obstruction
Solution Approach 1:
The algorithm incorporates feedback mechanisms at multiple stages: the detected aortic root position feeds into search volume definition, detected ostia positions feed into validation checks, and detection confidence metrics feed into iterative refinement. This feedback loop maintains high accuracy under obstruction by allowing the system to correct errors and reject unreliable detections.
Solution Approach 2:
The method performs preliminary localization of the aortic root and definition of a constrained search volume before attempting ostia detection. This preliminary action provides a reliable starting point that is less susceptible to obstruction artifacts, enabling automatic initialization to maintain high accuracy even in challenging cases.
4Measurement precision
If detailed aorta segmentation is performed before ostia detection, then detection can be achieved, but processing time increases
Solution Approach 1:
The method segments only the critical aortic root region rather than the entire aorta, dividing the problem into (1) rapid aortic root localization and (2) detailed ostia detection within a small search volume. This spatial segmentation reduces processing time by focusing computational resources only on the region containing the ostia.
Solution Approach 2:
The algorithm performs partial segmentation of only the aortic root and immediate surrounding region rather than complete aorta segmentation. This partial action is sufficient for detecting coronary ostia and significantly reduces processing time while maintaining the precision needed for accurate ostia localization.
Data Source
AI summary
A method and system for detection of native and bypass coronary ostia in a 3D volume, such as a CT volume, is disclosed. Native coronary ostia are detected by detecting a bounding box defining locations of a left native coronary ostium and a right native coronary ostium in the 3D volume using marginal space learning (MSL), and locally refining the locations of the left native coronary ostium and the right native coronary ostium using a trained native coronary ostium detector. Bypass coronary ostia are detected by segmenting an ascending aorta surface mesh in the 3D volume, generating a search region of a plurality of mesh points on the ascending aorta surface mesh based on a distribution of annotated bypass coronary ostia in a plurality of training volumes, and detecting the bypass coronary ostia by searching the plurality of mesh points in the search region.