Automatic Detection of Low-Quality Seismocardiogram Cycles Using the Outlier Approach

In this study, an algorithm was developed to automatically detect the low-quality (LQ) cardiac cycles in seismocardiogram (SCG). The proposed algorithm extracts some features from the SCG signal, which are referred to as signal quality indices (SQIs), and computes the outlier points of each SQI. Our hypothesis was that the identified cycles (outliers) would include the LQ ones. To verify this hypothesis, the algorithm results were compared with the LQ cycles that were labeled manually by an expert in the field. The developed algorithm was tested on total 1697 cardiac cycles, and there was a great overlap between the computed outliers and the LQ cycles (84% of 248 LQ cycles were identified). The proposed algorithm is simple, efficient, and works in an unsupervised manner.