Detection of aortic opening using ECG-free seismocardiography: Performance and limitations in aortic valve disease patients

Seismocardiography (SCG) is a non-invasive technique that measures chest movements associated with cardiac mechanical activity, including valve opening and closure. This study proposes and compares two automated methods for detecting aortic opening (AO) in SCG signals, and examines how aortic valve diseases (AVD) affect AO detectability without electrocardiogram (ECG) data. The relevance of systematic linear dorsoventral Z-axis (SCGz) selection was also compared with manual axis selection. The two proposed methods, based on successive variational mode decomposition (SVMD) and successive wavelet decomposition (SWD), were evaluated on two publicly available datasets and a new dataset comprising healthy and AVD subjects (in total 126 subjects). In healthy subjects (8,935 beats), SVMD achieved higher F1 scores than SWD (94.4 % vs 88.5 %) when systematically analysing the SCGz. In contrast, performance was more comparable between methods in aortic stenosis (16,955 beats), regurgitation (11,267 beats), and mixed AVD (2,414 beats) cases (65.2 % vs 64.1 %, 78.8 % vs. 76.9 %, and 81.4 % vs. 80.5 %, respectively). Both methods generally benefited from an axis selection based on expert consensus, compared to the usual systematic selection of SCGz, except in some patients with aortic regurgitation (AR). When considering the best F1 scores across the two approaches of axis selection, the results suggest that the SVMD-based method outperformed the SWD-based method in healthy subjects, whereas the differences were smaller in patients with AVD. Overall, these findings demonstrate the feasibility and the intrinsic limitations of ECG-free methods for detecting AO in the presence of waveform distortions induced by AVD.