Standalone Seismocardiogram-Based Detection of Aortic Stenosis and Mitral Regurgitation Using Variational Mode Decomposition

Aortic Stenosis (AS) and Mitral Regurgitation (MR), are significant contributors to cardiovascular morbidity and mortality. Automated methods are critical for early detection but often challenging with conventional methods. This work introduces a novel framework for simultaneously detecting AS and MR from Seismocardiogram (SCG) signals. The approach incorporates Variational Mode Decomposition (VMD) for signal analysis and introduces a standalone cardiac cycle extraction technique that segments SCG signals into individual cardiac cycles based on windowed energy of the decomposed modes. Seven spectral and statistical features are extracted from each cardiac cycle, providing comprehensive time-frequency representations of the signal. These features are input into XGBoost classifier, achieving a classification accuracy of 98.23%. The framework demonstrates the viability of SCG signals as a standalone, and non-invasive diagnostic modality for the screening of valvular heart diseases.