A research team has shown that the way people move can carry a surprising amount of information about their health. Using deep learning, the researchers built a model trained on three-dimensional skeletal motion data, and its learned representations were able to predict a large number of health traits across the body, according to a study published in Nature Medicine in November 2026.
The team worked with data from 3,328 adults who had been studied in considerable depth. Their movement was recorded with a depth camera while each participant performed five motor tasks, and the resulting motion data was used to train what the researchers describe as a gait foundation model. The approach treats movement as a rich signal that can be mined for patterns linked to physical and metabolic characteristics.
The results, as reported in the study, were broad. The learned embeddings predicted 1,887 of 3,210 phenotypic targets, meaning measurable traits that the researchers tested. In men, the model's predictions improved across all 18 body systems covered by the analysis, while in women the gains appeared in 17 of the 18 systems. The model also improved predictions of clinical diagnoses and of medication use, suggesting that movement data may reflect a person's overall medical profile and not only mobility.
The findings point toward movement analysis as a possible low-cost screening tool, since depth cameras are far less expensive and invasive than many laboratory tests. Yet the researchers and the reporting around the work leave important questions open. The materials reviewed here cover the abstract of the paper and a set of summary notes, not the full methods or validation details. It is also unclear how well the model would perform in populations different from the cohort studied, including people of other ages, backgrounds or health conditions.
Any move toward clinical use would require further testing in larger and more diverse groups before the approach could be relied upon. For now, the study adds to a growing body of research suggesting that the human body reveals much of its state through how it moves.
Credit. Based on reporting for Nature Medicine. Is this your work? Write to us.
2
