Decoding anxiety through your fingertip: mobile photoplethysmography-derived heart rate variability as an anxiety biomarker

• Main Author: Almeida, Ana Carolina Henriques
• Publication Date: 2024
• Format: Master thesis
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10773/44969
• topic: Photoplethysmogram; Heart rate variability; Anxiety; Mental health; Biomarker; Digital health

Anxiety is a global and worldwide concern, being reported as one of the scourges of the 21st century. The physiological response associated with anxiety is related to variations in the central and peripheral nervous system, whereas cardiovascular assessment is reported to be one of the potentially most informative areas in this domain. In particular, Heart rate variability (HRV) is one of the cardiovascular parameters widely studied as an anxiety biomarker. Ensuing, the ability to derive this biomarker from the photoplethysmography (PPG) signal acquired using a smartphone device would provide an easy and autonomous anxiety screening option, this is the case of the HOLI app by NEVARO. Notwithstanding, such mobile technologies present an insufficient validation methodology which is a major concern as it could lead to inappropriate screening. In this study, it was validated HOLI’s anxiety biomarking system through a multiphase process. First, due to its intended non-supervised acquisition, it was developed a novel PPG signal qualifier based on an adaptive threshold using the area of the blood volume pulse to mitigate poor signals. Furthermore, test results on a public database showed high sensitivities for binary classification, between noisy and clean signals, and the algorithm’s ability to identify noisy pulses accurately. In the next validation stage, a protocol was implemented on 47 subjects, within two locations, comprising the collection of the PPG and electrocardiography (ECG) signals during three states: Baseline, Physical Activation, and Psychological Activation (through emotion-inducing videos). The statistical analysis on HOLI’s cardiac estimation accuracy, showed large correlation and agreement values, except for the Physical Activation epoch, which presented high levels of motion artifacts. Concerning HRV anxiety biomarking functioning, the conjunction of statistical analysis with unsupervised learning methods demonstrated significance and expected behavior when using selected time and frequency domain HRV metrics for classifying anxiety subjectively. Further, these results support and extend past literature on the relationship between HRV and anxiety by introducing a novel HRV detection system with an embedded signal qualifier.