JMIR Publications

ABSTRACT

Background:

Accurate assessment of physical behaviours (PB) and energy expenditure (EE) is essential for public health research and digital health monitoring. Wearable accelerometers combined with machine learning (ML) or deep learning (DL) enable objective behaviour assessment, but most existing models are trained on laboratory data, limiting generalisability to free-living conditions.

Objective:

This study aimed to develop and validate multitask ML and DL models for PB classification across seven categories (sitting, standing, walking, running, sports, cycling, and lying) and EE across three intensity levels (sedentary, light, and moderate-to-vigorous physical activity) using thigh-worn (activPAL) and waist-worn (ActiGraph) wearable accelerometers. A second objective was to compare model-derived estimates of daily time spent in PB and EE categories across single- and dual-sensor (activPAL+ActiGraph) configurations, and to evaluate agreement between the best-performing model and corresponding estimates obtained from the proprietary activPAL CREA algorithm using free-living data collected over a 9-day monitoring period.

Methods:

Data were obtained from 590 adults in the multicentre WEALTH study (627 recruited) and included up to 9 days of concurrent activPAL and ActiGraph free-living recordings. Sparse accelerometer-labelled data were obtained using ecological momentary assessment and refined by retaining instances with ≥75% agreement with the CREA algorithm. Resulting labelled data of 583 participants were used to develop ML models for single-sensor (activPAL or ActiGraph) and combined (dual-sensor) configurations. A random forest (RF) model using engineered features and a multi-head convolutional neural network (MH-CNN) were trained within a multitask learning framework to jointly predict PB (task 1) and EE (task 2) using a subject-independent hold-out split. The test subset (n=87) was used to estimate daily time spent in PB and EE categories over 9 days, which were compared with CREA-derived estimates using Pearson coefficients and intraclass correlation coefficients (ICC).

Results:

The dual-sensor configuration consistently outperformed single-sensor models. For PB classification, the MH-CNN achieved the highest performance (F1-score 0.750). For EE, the RF model performed best (F1-score 0.741). Dual-sensor free-living estimates showed epidemiologically plausible distributions across the 24-hours period, including sitting 37% (538/1440 min), lying 34% (496/1440 min), walking 9% (131/1440 min) and moderate-to-vigorous intensity physical activity (MVPA) 2% (31/1440 min). Agreement with CREA was strongest for standing, walking, and cycling (r≥0.86; ICC≥0.72), while lying showed modest reliability (ICC=0.48). For EE, agreement was highest for light physical activity (LPA) and MVPA (ICC 0.72–0.75).

Conclusions:

Multitask models combining thigh- and waist-worn accelerometers, provide robust estimates of PB and EE under free-living conditions. The dual-sensor approach yielded more stable and epidemiologically coherent estimates than single-sensor methods, supporting its potential for large-scale population monitoring and mobile health applications.