Feasibility of heart rate monitoring in hemodialysis patients using visible imaging: a pilot study

Continuous vital signal monitoring is essential for patients undergoing hemodialysis sessions in which physiological instability, body fatigue, and cardiovascular complications may occur and require early medical interventions. However, conventional contactbased sensors such as electrocardiogram (ECG) have several limitations due to increased skin sensitivity, skin irritation, and interference with clinical procedures.

This study proposes a remote photoplethysmography (rPPG) framework for continuous heart rate monitoring of stage-5 chronic kidney disease (CKD-5) patients.

A pilot cohort of eight patients diagnosed with stage-5 chronic kidney disease undergoing hemodialysis sessions were monitored using synchronized RGB video and ECG reference measurements to generate 32 hours of video data. The proposed framework incorporates an automated facial landmark tracking and a dual-masking-based skin detection approach to robustly define regions-of-interest (ROI). A multimethod ensemble fusion is integrated through spectral analysis, peak detection, and harmonic product spectrum analysis for confidence-weighted final heart rate estimation.

The proposed framework achieved a mean absolute error (MAE) of 1.16 beats per minute (BPM) and root mean squared error (RMSE) of 1.86 BPM with a mean bias error of -0.68 BPM among all patients irrespective of lower visibility due to mask and partial occlusion.

These findings suggest that noncontact heart rate monitoring of patients diagnosed with advanced renal diseases is feasible using visible imaging and the rPPG method. By addressing clinical variability, the proposed framework suggests the potential of remote heart rate monitoring in a controlled dialysis-unit environment.