Radar Multiple Bin Selection for Breathing and Heart Rate Monitoring in Acute Stroke Patients in a Clinical Setting

Simple to use and accurate monitoring of stroke patients' breathing and heart rate in sleep is needed to lower the risk of secondary strokes and prevent worse functional outcomes due to disturbed sleep. In this study, we computed breathing and heart rates from clinical radar sleep recordings of 49 acute ischemic stroke patients. Parallel polysomnography served as the reference for evaluation. We compared radar rates computed using previously developed multiple and single range bin selection methods. Multiple selection yielded lower mean absolute errors (breathing: 0.39 vs. 0.87 breaths per min; heart rate: 0.84 vs. 3.99 beats per min) and higher correlations with the reference (breathing: 0.95 vs. 0.85; heart rate: 0.96 vs. 0.56). However, single range bin selection produced rates for a larger proportion of recording time (breathing: 93.49% vs. 73.38%; heart rate: 81.85% vs. 19.93%). Our results indicate that multiple range bin selection provides more accurate estimates of breathing and heart rate, but it has lower temporal coverage. Easy to use radar systems could facilitate the clinical adoption of contactless breathing and heart rate monitoring in sleep, improving the care provided to stroke patients.