Feasibility of unobtrusive camera-based heart rate extraction in a clinical interview scenario

In this study, we sought to evaluate the viability of camera-based pulse extraction in the context of clinical interviews involving individuals with affective disorders. Utilizing photoplethysmography imaging through five distinct methods, we compared the resulting heart rate with established photoplethysmogram reference signals. Our study, encompassing a substantial sample size of 140 participants, indicates patterns and trends in different subgroups by analyzing the heart rate extracted from the participants. Specifically, healthy controls had lower Mean Absolute Error scores (6.14Beats Per Minute overall) than individuals with major depressive disorder (8.24BPM) and schizophrenia spectrum disorder (9.36BPM). The complex relationship between MAE and exercise varied across these subgroups. This highlights the challenging interplay of factors that influence the relationship between photoplethysmography imaging and photoplethysmography measures in different clinical populations. In addition, we investigated the effects of movement during different phases of the clinical interview, with motion artifacts posing the greatest challenge to unobtrusive sensing. The study uncovered differences in movement patterns, particularly in tasks that required facial expressions. The findings demonstrated that decreased movement correlated with lower MAE scores across all subgroups, with the best MAE of 2.64BPM for healthy controls. This study broadens our understanding of remote pulse monitoring and emphasizes its potential applications and limitations in different clinical populations. The substantial sample size and careful analysis increase the robustness of our findings and provide valuable insights for future research and application of photoplethysmography imaging and unobtrusive monitoring of cardiac activity and vital signs in the clinical setting.