Abstract
Coronary angiography is the "gold standard" for diagnosing coronary artery disease (CAD). At present, the methods for detecting and evaluating coronary artery stenosis cannot satisfy the clinical needs, e.g., there is no prior study of detecting stenoses in prespecified vessel segments, which is necessary in clinical practice. Two vascular stenosis detection methods are proposed to assist the diagnosis. The first one is an automatic method, which can automatically extract the entire coronary artery tree and mark all the possible stenoses. The second one is an interactive method. With this method, the user can choose any vessel segment to do further analysis of its stenoses. Experiments show that the proposed methods are robust for angiograms with various vessel structures. The precision, sensitivity, and $F_1$ score of the automatic stenosis detection method are 0.821, 0.757, and 0.788, respectively. Further investigation proves that the interactive method can provide a more precise outcome of stenosis detection, and our quantitative analysis is closer to reality. The proposed automatic method and interactive method are effective and can complement each other in clinical practice. The first method can be used for preliminary screening, and the second method can be used for further quantitative analysis. We believe the proposed solution is more suitable for the clinical diagnosis of CAD.
Introduction
Coronary artery disease (CAD) remains a leading cause of mortality worldwide. While Coronary Angiography (CA) is the definitive diagnostic tool, its interpretation is often subjective, time-consuming, and prone to human error. Existing computer-aided diagnosis (CAD) tools often lack the quantitative precision required for surgical planning or fail to allow clinicians to focus on specific high-risk vessel segments. This paper introduces a hybrid approach combining an automated screening algorithm with a precise, user-driven interactive tool to bridge the gap between computer vision and clinical utility.
Methodology
The proposed system relies on robust image preprocessing, including ROF denoising and Contrast Limited Adaptive Histogram Equalization (CLAHE), to prepare grayscale coronary angiograms for analysis.
1. Automatic Stenosis Detection
The automatic method utilizes an Improved Adaptive Geometrical Tracking Algorithm (IAGVT). It begins by extracting the vessel centerline using the Hessian matrix to identify "ridge points" and determine local vessel direction.
A critical challenge in automated tracking is handling vessel bifurcations. The IAGVT addresses this by searching a "fan ring" area to identify new branches, enabling the algorithm to map the entire arterial tree autonomously.
To refine the tracking, the Chan-Vese (C-V) model is integrated. This region-based active contour method adjusts tracking points to ensure they remain centered, even in images with uneven grayscale distributions.
2. Interactive Stenosis Detection
The interactive method is designed for detailed analysis of specific vessel segments. A clinician selects a start and end point, and the system tracks the segment using a novel Potential Energy Function.
This function acts as a "gravitational pull," rewarding the algorithm for staying within the vessel boundaries while simultaneously pulling it toward the target endpoint. This mechanism effectively prevents the "tracking drift" common in older algorithms when encountering complex overlaps or loops.
Experimental Results
The methods were evaluated using a clinical dataset of 53 grayscale coronary angiograms from 30 patients, provided by Qilu Hospital. Expert manual annotations served as the ground truth.
The automatic method demonstrated superior performance compared to established baselines, achieving a Precision of 0.821, a Sensitivity of 0.757, and an F1 score of 0.788.
The interactive method provided even higher precision, successfully identifying subtle narrowings missed by the automated system. Quantitative analysis revealed that the relative error in vessel diameter measurement was significantly reduced, providing a "stenotic degree" calculation that closely aligns with physical reality.
Conclusion
By combining automated global screening with targeted interactive analysis, this dual-method system offers a robust solution for CAD diagnosis. The automatic method serves as an efficient preliminary screening tool, while the interactive method provides the high-precision quantitative data necessary for clinical decision-making. This hybrid approach enhances diagnostic accuracy and better aligns with the practical needs of cardiologists.
Summary updated