Towards Wearable Acute Stress Detection and Mitigation via Real-Time Photoplethysmogram Feature Detection

Farhan N. Rahman, Prabhkirat S. Bindra, Jesus Antonio Sanchez-Perez, Afra Nawar, Harrison Trask Crane, John A. Berkebile, Asim Hossain Gazi, Jin-Oh Hahn, Omer T. Inan

Abstract: Acute stress has been linked to an increased risk for adverse cardiac events, necessitating the development of techniques for continuous stress monitoring and regulation. Wearable sensing can be used to continuously monitor cardiovascular signals relevant to acute stress. However, to enable real-time stress mitigation techniques in a wearable system, both cardiac and vascular features of acute stress need to be extracted from these signals in real-time. We have developed an efficient algorithm capable of accurately extracting heart rate (HR) and photoplethysmogram (PPG) amplitude (PPGamp) from PPG signals on resource-constrained wearable devices. This algorithm was deployed on a custom-designed wearable device and was compared against a reference electrocardiogram and standard signal processing techniques in a protocol involving acute stressors. The device consumed an average of 113 mA during the protocol. Through correlation and Bland-Altman analyses, we found that the features extracted using our algorithm were significantly $(p < 0.001)$ correlated (HR: $r=0.919$, PPGamp: $I {=0.993}$) and strongly agreed with (HR: $< 5$ beats per minute difference, PPGamp: $< 5\%$ difference at 95% limits of agreement) those derived using the benchtop devices and post-processing algorithms. These results demonstrate our algorithm's ability to extract both the cardiac and vascular effects of acute stress in real-time, enabling future work in the area of wearable closed-loop stress detection and mitigation.