SIGNAL-QUALITY-AWARE ECG–PPG FUSION FOR EARLY FLAGGING OF SIMULATED HEMODYNAMIC INSTABILITY DURING HEMODIALYSIS
Abstract
Intradialytic hemodynamic instability is a clinically important challenge during hemodialysis, but reliable automated detection from wearable biosignals is complicated by motion artifacts, inter-patient variability, and signal-quality changes. This paper presents a lightweight ECG–photoplethysmography (PPG) feature-fusion pipeline designed to flag candidate instability windows while explicitly incorporating a PPG signal-quality proxy. A five-minute synthetic experiment was generated at 100 Hz with an ECG-like pulse train, a PPG-like waveform, a simulated event interval, baseline drift, random noise, and transient artifact bursts. Ten-second windows with two-second overlap were represented by heart-rate proxy, PPG amplitude change, and a residual-based signal-quality feature. A weighted score was thresholded using the 95th percentile of the simulated baseline period. On this single deterministic simulation, the fused detector achieved sensitivity 0.900, specificity 0.952, and precision 0.878. These figures describe only the constructed synthetic scenario; they are not clinical performance estimates. The contribution is a transparent software-oriented baseline for future evaluation on ethically approved, annotated hemodialysis recordings.