Abstract 17703: Sequential Validation of a Cuffless Photoplethysmography-Based Wristband for Blood Pressure Measurement: Algorithm Building to Regulatory Standard Trial

Abstract

Introduction: Hypertension increases cardiovascular risk. Measuring reliable blood pressure (sBP and dBP for systolic and diastolic BP) is challenging. A photoplethysmography (PPG)-based wristband with a custom BP algorithm could provide continuous BP. The algorithm was made by PPG channel aggregation, band-pass filtering and segment creation and evaluated the different PPG colors. Noise and interference were removed. PPG segments were created with a quality factor based on pulse quality. The AI model learned patterns from structured data, including pre-processed features and initialization measurements, ranking their importance. Hypothesis: cuffless PPG-based-wristband method for continuous BP monitoring is compatible with ISO 81060-2:2019. Methods: Compare PPG-guided BP algorithm predictions with subclavian arterial reference measurements during cardiac catheterization. Consecutive patients meeting ISO 81060-2:2019 criteria were included. Reference measurements used a validated invasive BP device (100Hz). PPG signals were collected at 128Hz using six light emission diodes and two photodiodes. Three initialization measurements with a validated cuff were taken before catheterization. Machine learning-based BP algorithm utilized 100+ features. Correlation, mean error, and standard deviation (SD) and sBP and dBP were determined. Results: 97 patients provided 420 samples. Mean age, weight, and height were 67 years, 183.7 lbs, and 5'8.5" respectively. sBP swas ≤100mmHg (11%) and ≥160mmHg (25%). dBP was ≤70mmHg (53%) and ≥85mmHg (24%). BP algorithm predictions correlated strongly with reference measurements for sBP (r = 0.985) and dBP (r = 0.961) BP. Mean error was ±3.7mmHg (SD 4.4 mmHg) for sBP and ±2.5mmHg (SD 3.7 mmHg) for dBP. Results were consistent across gender and skin color categories (Fitzpatrick I-VI), but also for different strata of BP values. Conclusions: Wristband-based PPG with the developed BP algorithm provides accurate continuous BP monitoring across various BP ranges. It offers a valid and less burdensome alternative to cuff BP measurements for in-hospital and at-home monitoring. Further research is needed to evaluate algorithm precision during movement and for long-term prediction stability.