AI Learns to Detect Signs of Nearly 1,900 Health Indicators from a Person's Gait
20:43 10 November, 2026Artificial intelligence can identify signs of various diseases by analysing a person's movement patterns, including their gait. This is the conclusion reached by researchers whose study was published in the journal Nature Medicine.
The model developed by the researchers identified statistically significant associations between movement patterns and 1,887 of 3,210 medical indicators covering 18 body systems.
The study used data from 3,328 adult participants in the Human Phenotype Project. The researchers analysed 17,589 movement sequences recorded by cameras during five standardised movement tests. The AI model identified patterns in three-dimensional skeletal movement data that conventional measurements, such as walking speed and stride length, may fail to capture.
Associations were found with indicators of cardiovascular and metabolic health, bone density, liver function, sleep, mental health and physical frailty. Moreover, movement analysis improved predictions even after accounting for age, body mass index, visceral fat levels and height.
One of the most notable findings was the association between gait and liver health. The model predicted liver stiffness, a measure used to assess fibrosis – the formation of scar tissue – non-invasively. In some cases, adding movement data improved predictions as much as using standard blood tests to assess liver function.
The AI also identified patterns associated with depression and metabolic disorders. In a screening analysis, adding movement data increased the number of detected cases of depression among men by 81% and diabetes cases by 83%.
According to the researchers, different parts of the body provide information about different aspects of health. Lower-body movements proved particularly informative for assessing metabolic health and physical frailty, while torso movements were more useful for predicting sleep and lifestyle indicators.
The technology could be used in clinical drug trials. Traditional tests often reduce a patient's physical capabilities to a single measure, such as the time taken to walk a certain distance or a score on a functional assessment scale.
A more detailed analysis of movement makes it possible to assess coordination, balance, stability and the effort required to walk, as well as distinguish genuine functional recovery from compensatory movements.