Evaluation of ASM for Ventricular Segmentation in Patients with Diverse Cardiac Abnormalities

The efficacy of Active Shape Models (ASM) for automated ventricular segmentation was evaluated to address the computational demands of manual segmentation and the interpretability limitations of deep learning. A statistical shape model was constructed using a limited cohort of 19 Coronary Computed Tomography Angiography (CCTA) scans derived from patients with diverse cardiac abnormalities. Principal Component Analysis (PCA) was employed to encapsulate morphological variability, and strict point correspondence was enforced to maintain topological consistency. Validation was conducted via leave-one-out cross-validation, benchmarking automated segmentations against expert-delineated ground truths using the Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD). It was found that mean Dice scores ranged from 0.50 to 0.64, with individual high-fidelity cases achieving scores up to 0.84. These results indicated that while quantitative performance reflected the complexity of pathological morphology, the methodology successfully accommodated high morphometric variance. It can be concluded that the ASM framework provides a resilient, interpretable foundation for managing complex clinical geometry where unconstrained models may fail.