Publicly available signal databases containing seismocardiographic signals — the state in early 2023
Executive Summary
This study reviews the state of publicly available seismocardiographic (SCG) signal databases as of early 2023, identifying five datasets with varying subject demographics and experimental conditions. The datasets include signals from healthy individuals, patients with valvular heart disease (VHD), and porcine models of hemorrhage, with accompanying reference signals such as ECG, GCG, and PPG. Key limitations include gender bias, imbalance between healthy and diseased subjects, and limited representation of cardiovascular conditions, which may hinder the development of robust classification models.
Answer Machine Insights
Q: What are the main limitations of the SCG datasets reviewed?
The datasets exhibit gender bias, imbalance between healthy and diseased subjects, and limited representation of cardiovascular conditions.
The limitations of the currently available SCG signal databases are gender biased toward male subjects, there is no representation of cardiovascular conditions other than VHD and hemorrhage, and the imbalance between the number of signals from healthy subjects and patients with cardiovascular conditions.
Q: What is the purpose of the CEBSDB dataset?
The CEBSDB dataset was created to study the influence of breathing on heartbeat detection and compare cardiac intervals derived from ECG and SCG signals.
The purpose of recording the seismocardiograms in CEBSDB was to check the influence of slight errors in heartbeat detection that could be caused by breathing and to compare the cardiac intervals derived from ECG and SCG signals.
Key Results
Five publicly available SCG datasets were identified, including signals from 62 healthy subjects and 106 subjects with cardiovascular conditions.
Gender bias was observed, with 116 male subjects compared to 52 female subjects across all datasets.
Clinical Snapshot
Evidence Rating
Relevance
high Priority