Detecting Coronary Artery Disease Using Rest Seismocardiography and Gyrocardiography

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Coronary artery disease (CAD) diagnosis typically relies on invasive or stress-based methods.
• There is a need for non-invasive, accessible tools for early CAD detection.
• Seismocardiogram (SCG) and gyrocardiogram (GCG) offer potential physiological signals for cardiac assessment.

Purpose of the Study:

• To develop and validate a non-invasive solution for identifying coronary artery disease (CAD).
• To assess the efficacy of seismocardiogram (SCG) and gyrocardiogram (GCG) signals in predicting CAD risk.
• To compare the performance of SCG/GCG-based models against traditional stress electrocardiography (ECG).

Main Methods:

• Collected SCG and GCG data from 310 individuals across multiple centers using an accelerometer/gyroscope sensor.
• Extracted time-frequency features from SCG and GCG signals.
• Trained six separate 1D Convolutional Neural Network (1D CNN) classifiers and fused their results to estimate CAD risk.
• Validated predictions against coronary angiography results.

Main Results:

• SCG-based classifiers, particularly those using the z and y axes, demonstrated high performance with an AUC of 91% (p < 0.05).
• SCG models achieved superior sensitivity (92-94%) compared to GCG models (73-87%) for CAD detection.
• The 3-axial SCG/GCG resting-state solution performed comparably to or better than stress ECG, without improvement from combining all SCG/GCG classifiers.

Conclusions:

• SCG signals provide a highly effective, non-invasive method for predicting coronary artery disease risk.
• The proposed 3-axial SCG/GCG system shows promise as a portable, at-home screening tool for CAD.
• This technology could potentially reduce reliance on more invasive or resource-intensive diagnostic procedures.