Novel Wearable Seismocardiography and Machine Learning Algorithms Can Assess Clinical Status of Heart Failure Patients
Executive Summary
This study utilized a wearable patch equipped with ECG and seismocardiogram sensors to assess heart failure (HF) states in patients during a 6-minute walk test (6MWT) and recovery. Using graph mining techniques to compute a graph similarity score (GSS), the researchers demonstrated significant differences between compensated and decompensated HF patients, with GSS correlating to clinical improvement. The findings suggest that wearable technologies combined with machine learning can noninvasively monitor HF progression and response to treatment.
Answer Machine Insights
Q: What metric was used to differentiate compensated and decompensated HF patients?
The graph similarity score (GSS) derived from seismocardiogram signals was used.
We found that graph similarity score can assess HF patient state and correlates to clinical improvement in 45 patients (13 decompensated, 32 compensated).
Q: How does the wearable patch measure cardiac function?
The patch uses a triaxial accelerometer to capture seismocardiogram signals representing chest vibrations caused by heart and blood movement.
The SCG signal represents the vibrations of the chest wall in response to the movement of the heart and blood with each heartbeat.
Key Results
Graph similarity score (GSS) was significantly higher in decompensated HF patients compared to compensated patients (44.4±4.9 vs. 35.2±10.5; P<0.001).
Longitudinal data showed a significant decrease in GSS from admission to discharge in decompensated HF patients (44±4.1 vs. 35±3.9; P<0.05).
Visual Evidence
![Figure 2. General overview of the proposed method. A, The seismocardiogram (SCG) signals (3 axes [dorso-ventral, DV, head-to-foot, HF, and lateral, LAT]) of heart are measured using the custom, wearable patch. Filtering is applied and SCG3D (a point-by-point average of the 3 axes) is calculated. Then, SCG3D of rest and recovery segments are windowed and generate NRES and NREC. B, The L frequency domain feature sets (X and Y) are computed from NRES and NREC. C, Two k-nearest neighbor graphs (GX and GY) are constructed from frequency feature vectors X and Y , and GSS is calculated to measure similarity in between rest and recovery states. D, An example of graph similarity score (GSS) calculation between 2 illustrative graphs. 6MWT indicates 6-minute walk test.](/_next/image?url=%2Fscience-hub%2Fimages%2F10.1161_circheartfailure.117.004313%2F10.1161_circheartfailure.117.004313_p5_i1.png&w=3840&q=75)
Figure 2. General overview of the proposed method. A, The seismocardiogram (SCG) signals (3 axes [dorso-ventral, DV, head-to-foot, HF, and lateral, LAT]) of heart are measured using the custom, wearable patch. Filtering is applied and SCG3D (a point-by-point average of the 3 axes) is calculated. Then, SCG3D of rest and recovery segments are windowed and generate NRES and NREC. B, The L frequency domain feature sets (X and Y) are computed from NRES and NREC. C, Two k-nearest neighbor graphs (GX and GY) are constructed from frequency feature vectors X and Y , and GSS is calculated to measure similarity in between rest and recovery states. D, An example of graph similarity score (GSS) calculation between 2 illustrative graphs. 6MWT indicates 6-minute walk test.
Clinical Snapshot
Evidence Rating
Relevance
high Priority