Template Generation and Fiducial Point Annotation for a Seismocardiographic Signal Acquired Near the Mitral Valve

In previous studies, seismocardiography (SCG) signals have been acquired from multiple locations on the chest wall. However, SCG near the mitral valve (MV) has not been investigated, and its utility in heart disease diagnosis has not been evaluated. This study assessed SCG applied to the clinical auscultation site of the mitral heart sound and generated SCG templates with annotated fiducial points. The clinical measurements of SCG and echocardiography were used for analysis. The flowchart elucidated analysis approaches based on training set and testing set division, signal processing paths for fiducial point identification, and template generation and validation. A morphological analysis leveraged the dynamic time warping (DTW) algorithm to align SCG signals with different heart rates (HRs). Myocardial contractility indexes were derived by calculating cardiac timing intervals obtained from SCG fiducial points. The resultant cardiac indexes obeyed the negative proportional relationship of the left ventricular ejection fraction (LVEF) with a coefficient of determination of
$\ge 0.75$
, and the tests of receiver operating characteristic (ROC) curve prediction demonstrated a minimum area under the ROC curve score of 91.2% in identifying patients with heart failure (HF) with reduced ejection fraction. In this study, MV SCG was proven to be applicable for heart disease diagnosis for the first time. The cardiac index thresholds obtained in the experiments could be evaluated as surrogates of LVEF criteria for the assessment of myocardial contractility. This technique is expected to be used to reduce the readmission rate of patients with HF or those with ventricular dyssynchrony-related diseases.