Enhancing Fiducial Point Detection in Imaging Photoplethysmography Through Spatial Fusion

Imaging photoplethysmography (iPPG) enables contactless assessment of blood volume changes but is susceptible to noise, particularly under impaired perfusion. This work proposes a spatial fusion approach that weights local iPPG signals based on signal quality metrics (SNR, Perfusion Index, Signal Quality Index) to preserve waveform morphology. The method is evaluated on a dataset comprising synchronized iPPG and reference PPG recordings from 20 healthy participants under baseline, light, and moderate cuff-induced perfusion impairment. Detection performance is quantified using the detection rate (DR) of fiducial points (FP). Results demonstrate a significant improvement in DR with spatial fusion compared to conventional spatial averaging. Under moderate impairment, gains of up to 8 percentage points are observed for dicrotic notch and diastolic peak detection (p < 0.01), while onset and systolic peak detection remain unaffected.