An open-source deep learning model for classifying real-world activity intensities in persons with chronic pain using a wrist-worn accelerometer

Introduction: In chronic pain (CP) care, patterns of activity and rest are relevant across diagnostic, intervention, and evaluation phases. Objective measurement of these patterns can support individualized care. Wrist-worn tri-axial accelerometers are user-friendly and yield valid and reliable data in the general population. Public availability of a detailed dataset together with a validated deep learning (DL) model for classifying activity intensities from raw accelerometer data can promote clinical use and standardize activity intensity measurement in CP research. Methods: This study aimed to (1) develop, test, and share a DL model that classifies activity intensity from raw accelerometer data, (2) explore whether CP-specific training improves performance in CP, and (3) explore construct validity in real-world settings among individuals with and without CP, using real-world observations as comparator instrument. Seven strategies were applied to train a Convolutional Neural Network, using three datasets comprising 211,864 ten-second windows. Construct validity was evaluated using 3,443 one-minute real-world observations from 5 individuals with and 10 without CP. Results: Classification accuracy was 76.5% in individuals with CP and 80.5% in individuals without CP. There was no clear advantage of CP-specific training. Real-world agreement was 73.6% in CP and 78.6% in individuals without CP. Discussion: Lower and individually variable real-world agreement, particularly in CP, indicates that model performance may depend on individual characteristics and the comparator instrument. The results suggest that the accompanying script and datasets enable accurate processing of raw accelerometer time series to activity intensities in the CP populations studied. Further investigation of construct validity and other clinimetric properties is warranted.