Potential SCG Predictors of Heart Failure Readmission

Introduction Recent advances in sensor technology, digital processing and artificial intelligence can be used to potentially increase the utility of seismocardiography (SCG) for ambulatory assessment of heart failure (HF) patients. SCG respiratory variability, energy spectrum shifts to ultra-low frequencies, third heart sound amplitude and other features may provide early warning of HF deterioration. If these non-invasively detected changes prove to be reliable, then earlier intervention may reduce readmissions, costs, and perhaps morbidity and mortality. Objective The objective of this study was to investigate early SCG waveform features that may be predictive of HF re-admission. Methods SCG was longitudinally followed in 40 HF patients, 9 (23%) of whom were readmitted during the study period (median follow-up 1.5 mos). SCG waveforms were clustered into two groups (that correlate with respiration) using unsupervised machine learning. Following this, various signal features were calculated using digital signal processing including waveform respiratory variability, sub-audible spectral energy distribution, and SCG3 (“S3”) amplitude. Results Similar to prior observations with heart rate variability, the intra group SCG variability tended to decrease as the patient approached re-admission (table below). The energy ratio between the ultra-low frequency band (0.5-5 Hz) and the broader sub-audible frequency band (0.5-15 Hz) showed a shift towards the lower frequencies as the subjects approached re-admission. In addition, SCG3 amplitude (which corresponds to S3) appeared to increase as the subjects approached readmission. Conclusions We report preliminary data showing some SCG features that appeared to be early predictors of HF readmission. Features included Intra-group variability, along with the SCG downward frequency shifts and increased SCG3 amplitude Ongoing work will help to better characterize and validate these initial observations