Our team built a deep learning model called CAFusionNet-MC, a modification of the CAFusionNet classification architecture
(Li et al. 2025) applied to cardiac phonocardiogram (PCG) analysis. Inspired by Samarappuli et al. 2026, we implemented a Markov
Chain (MC) refinement engine with three different algorithms to augment abnormal data before applying CAFusionNet to address
the critical scarcity of abnormal cardiac data. This hybrid approach allows CAFusionNet to further capture the characteristics of heart
diseases hidden in cardiac acoustic data, and provides a high-fidelity, user-configurable solution for augmenting imbalanced clinical datasets.