X-ECGNet: An Interpretable DL model for Stress Detection using ECG in COVID-19 Healthcare Workers

Abstract

COVID-19 pandemic erupted in December 2019, spreading extremely fast and stretching the healthcare infras-tructure of most countries beyond their capacities. This impacted the healthcare workers (HCW) adversely because 1) they were pressured to work almost round the clock without a break; 2) they were in close contact with the COVID-19 patients and hence, were at high risk; and 3) they suffered from the fear of spreading COVID to their families. Hence, many HCWs were stressed and burnout. It is known that stress directly affects the heart and can lead to serious cardiovascular problems. Currently, stress is measured subjectively via self-declared questionnaires. Objective markers of stress are required to ascertain the quantitative impact of stress on the heart. Thus, this paper aims to detect stress contributing factors in HCWs and determine the changes in the ECG of stressed HCWs. We collected data from multiple hospitals in Northern India and developed a deep learning model, namely X-ECGNet, to detect stress. We also tried to add interpretability to the model using the recent method of SHAP analysis. Deployment of such models can help the government and hospital administrations timely detect stress in HCWs and make informed decisions to save systems from collapse during such calamities.