Remote COPD Severity and Exacerbation Detection Using Heart Rate and Activity Data Measured from a Wearable Device
Abhishek Tiwari, Salaar Liaqat, Daniyal Liaqat, Moshe Gabel, Eyal de Lara, Tiago H. Falk
Abstract
Chronic obstructive pulmonary disease (COPD) is one of the leading causes of human mortality worldwide. Traditionally, estimating COPD severity has been done in controlled clinical conditions using cough sounds, respiration, and heart rate variability, with the latter reporting insights on the autonomic dysfunction caused by the disease. Advancements in remote monitoring and wearable device technologies, in turn, have allowed for remote COPD monitoring in daily life conditions. In this study, we explore the potential for predicting COPD severity and exacerbation using a low-cost wearable device that measures heart rate and activity data. We collected smartwatch sensor data from 35 COPD patients over a period of three months. Our evaluation shows that future trajectory of the disease can be predicted using only the first few days of continuous unobtrusive wearable data collected from COPD patients. Using features extracted from wearable device an Isolation Forest was able to predict exacerbation with an area under curve (AUC) 0.69 thus showing improvement over a random choice classifier.