Comprehensive Analysis of Cardiogenic Vibrations for Automated Detection of Atrial Fibrillation Using Smartphone Mechanocardiograms
Executive Summary
This study presents a machine learning-based approach for detecting atrial fibrillation (AFib) using smartphone-derived seismocardiography (SCG) and gyrocardiography (GCG) signals. The algorithm was validated on a dataset of 435 subjects, achieving accuracy, sensitivity, and specificity of up to 97%, 99%, and 95% in cross-validation, and 95%, 93%, and 97% in cross-database testing. The findings demonstrate the feasibility of self-monitoring AFib using widely available smartphones without additional hardware, offering a cost-effective solution for large-scale screening.
Answer Machine Insights
Q: What is the diagnostic performance of the smartphone-based AFib detection algorithm?
The algorithm achieved accuracy, sensitivity, and specificity of up to 97%, 99%, and 95% in cross-validation, and 95%, 93%, and 97% in cross-database testing.
The experimental results showed accuracy, sensitivity, and specificity of approximately 97%, 99%, and 95% for the CV study and up to 95%, 93%, and 97% for the CD test, respectively.
Q: What is the significance of combining SCG and GCG signals?
Combining SCG and GCG signals significantly improved the classification performance, achieving near-perfect agreement with visual interpretation of telemetry ECG recordings.
The Cohen’s kappa coefficient of the SCG-GCG features was 0.94 indicating a near-perfect agreement in rhythm classification between the smartphone algorithm and visual interpretation of telemetry ECG recordings.
Key Results
Cross-validation study achieved 97% accuracy, 99% sensitivity, and 95% specificity.
Cross-database testing achieved 95% accuracy, 93% sensitivity, and 97% specificity.
Clinical Snapshot
Evidence Rating
Relevance
high Priority