Detection of Left Ventricular Ejection Fraction Abnormality Using Fusion of Acoustic and Biopotential Characteristics of Precordium

This study develops a wearable monitoring platform for the detection of abnormal left ventricular ejection fraction (LVEF) using a fusion of an accelerometer contact microphone (ACM) and an electrocardiogram (ECG) sensor. Two signal processing chains are designed to annotate ACM and ECG recordings. Afterwards, the pre-ejection period (PEP) and left ventricular ejection time (LVET) are estimated as the time difference between the first heart sound (S
1
) and the R-peak in ECG signals, and the time difference between the first and second heart sounds (S
1
and S
2
), respectively. The ratio of PEP to LVET is then utilized to differentiate between healthy and abnormal-LVEF groups. The model is evaluated on 15 subjects (8 healthy subjects and 7 subjects with LVEF abnormality) where the ground truth values are the LVEF parameter acquired by the echocardiography machine. An average (± standard deviation) accuracy of 84.47% (± 17.58%) is obtained for the detection of LVEF abnormality for a total of 5989 heartbeats. It is demonstrated that the proposed method is capable of LVEF abnormality detection with accuracies within the range of 54.35% - 100%.