Boosted-SpringDTW for Comprehensive Feature Extraction of PPG Signals

Abstract

To achieve high-quality comprehensive feature extraction from physiological signals that enables precise physiological parameter estimation despite evolving waveform morphologies. Methods: We propose Boosted-SpringDTW, a probabilistic framework that leverages dynamic time warping (DTW) and minimal domain-specific heuristics to simultaneously segment physiological signals and identify fiducial points that represent cardiac events. An automated dynamic template adapts to evolving waveform morphologies. We validate Boosted-SpringDTW performance with a benchmark PPG dataset whose morphologies include subject-and respiratory-induced variation. Results: Boosted-SpringDTW achieves precision, recall, and F1scores over 0.96 for identifying fiducial points and mean absolute error values less than 11.41 milliseconds when estimating IBI. Conclusion: Boosted-SpringDTW improves F1-Scores compared to two baseline feature extraction algorithms by 35% on average for fiducial point identification and mean percent difference by 16% on average for IBI estimation. Significance: Precise hemodynamic parameter estimation with wearable devices enables continuous health monitoring throughout a patients' daily life.

Pulse2AI: An Adaptive Framework to Standardize and Process Pulsatile Wearable Sensor Data for Clinical Applications

“…Subsequently, we used temporal clustering to cluster cardiac cycles based on associated latent features with Gaussian mixture model (GMM) [34] . With an empirically selected “gold standard” template denoting an ideal and clean cardiac cycle [35] , we considered clusters whose centroids have lower normalized Dynamic Time Warping (DTW) distance to be cleaner. Whereas clusters whose centroids with normalized DTW distance surpassing a defined threshold d are rejected.…”

Continuous cuffless blood pressure monitoring with a wearable ring bioimpedance device

“…We quantified the amount of deviation in beat-to-beat bioimpedance morphology using three different metrics: (1) standard deviation (SD) from the mean waveform morphology, (2) signal-to-noise ratio (SNR) from the fundamental heart rate frequency to the noise floor measured through the Fast Fourier Transform (FFT) of the full 30seconds cycle, and (3) dynamic time warping (DTW) distance of waveforms to the mean waveform morphology for every beat (see Methods). We chose DTW distance as a metric to measure waveform similarities instead of Euclidean distance to account for morphological changes in waveforms from beat-to-beat due to the physiological reasons (e.g., blood flow rate impacting the timing of step (iii)-reflection wave arrival) 28 . Figure 5a provides an example for the ensembled bioimpedance beats for different ring separations and configurations with signals acquired from the same participant, where smaller ring electrode separation of Configuration 1 shows a more consistent waveform pattern compared to the case of higher separation or different configurations of injection and sensing electrode connections.…”

Noninvasive blood oxygen, heartbeat rate, and blood pressure parameter monitoring by photoplethysmography signals

“…Based on the various factors discussed above, we choose the relevant literature on the systolic and diastolic blood pressure calculation formulas derived from the waveform characteristic points in the PPG time domain with a partial theoretical basis as the starting point for this study [ 18 , 19 ]. First, the self-developed hardware circuits, firmware, and software programs to calculate blood pressure are reviewed.…”