From video to vital signs: a new method for contactless multichannel seismocardiography
Executive Summary
This study introduces a novel, cost-effective method for multichannel seismocardiography (SCG) using smartphone video recordings and deep learning techniques. By employing a grid of QR-coded stickers on the chest, the researchers utilized YOLOv7 for object detection and template tracking to extract SCG signals, achieving high-resolution multichannel SCG maps. The method demonstrated strong agreement with gold-standard ECG-derived heart rate (HR) measurements (bias ± 1.96 SD = 0.04 ± 2.14 bpm; r = 0.99, p < 0.001) and identified optimal chest locations for HR estimation. This approach has significant potential for accessible, non-invasive cardiac monitoring using widely available smartphones.
Answer Machine Insights
Q: How accurate is the heart rate estimation using this method compared to ECG?
The heart rate estimation showed a bias ± 1.96 SD of 0.04 ± 2.14 bpm compared to ECG-derived HR, with a correlation coefficient of r = 0.99 (p < 0.001).
Moreover, our algorithm accurately estimated HR from 1968 SCG signals extracted from the videos compared to the gold-standard HR measured from each subject’s electrocardiogram (bias ± 1.96 SD = 0.04 ± 2.14 bpm; r = 0.99, p < 0.001).
Q: What is the significance of using multichannel SCG over single-channel SCG?
Multichannel SCG enhances spatial resolution, enabling precise identification of optimal chest locations for physiological measurements and providing a more comprehensive view of cardiac activities.
This multichannel capability allows for the construction of SCG maps that provide a more comprehensive view of cardiac activities, including rapid ejection phases and valve dynamics, which were not as detailed in our previous work.
Key Results
The proposed method achieved a mean squared error of 0.1078 and 0.0418 for right-to-left and head-to-foot SCG signals, respectively.
Heart rate estimation from SCG signals showed a bias ± 1.96 SD of 0.04 ± 2.14 bpm compared to ECG-derived HR, with a correlation coefficient of r = 0.99 (p < 0.001).
Visual Evidence
Clinical Snapshot
Evidence Rating
Relevance
high Priority