Classification of Heart Failure Based on Phase Trajectory Complexity of Seismocardiogram

Objective: Heart failure (HF) is a leading cause of disease progression and mortality in HF patients. This study aimed to classify HF by extracting nonlinear features from the seismocardiogram (SCG) signal of heart. First, the trajectory of SCG signal was depicted in a reconstructed high-dimensional phase space to reflect nonlinear properties and cardiac dynamics of SCG. This trajectory was then mapped onto a “displacement-velocity” phase plane to obtain the 2-D phase plane trajectory. Thereby, a phase trajectory complexity (PTC) of the SCG signal based on entropy was calculated and proposed as a key feature in order to represent the complexity of the cardiac dynamic system and realize HF classification. The available SCG signals of 73 HF patients and 20 healthy individuals used in this study were obtained from the SCG-right heart catheterization (RHC) and combined measurement of ECG, breathing, and SCGs databases, respectively. Experimental results indicate that our proposed features achieve an accuracy of 87.68% in HF classification. When combined with other nonlinear features, the highest training and prediction accuracies reach 100% and 95.17%, respectively. This demonstrates that our proposed features contribute to improving the HF classification outcomes.