Walking towards the future: Exploring OpenTUG's validity in automatic walking activity analyses and the relationship...

Vestibular disorders impair body orientation and spatial navigation, often causing cognitive and affective symptoms, reducing quality of life. Traditional assessments involve time-consuming physiological and neuropsychological evaluations. This study investigates the feasibility and validity of applying a 2D human pose estimation (OpenTUG) to automatically analyze the walking activity of patients with vestibular disorders. Forty-eight patients undergoing treatment for vestibular disorders and 26 mean age-matched controls performed the TUG at normal, slow, and fast walking speeds; neuropsychological evaluations assessing attention, and quality-of-life questionnaires. Patients required significantly more time and steps to complete the TUG than controls at normal and slow speeds (e.g., normal speed: 7319 ms vs. 5976 ms, p = .002). Automated OpenTUG and manual analyses were highly correlated (time: all r > .94, p < .001; steps: all r > .74, p < .001). Patients reported higher dizziness-related handicap (median DHI total: 30 vs. 0, p < .001) and psychological symptoms (median HADS total: 12 vs. 8, p = .005). Exploratory analyses revealed moderate correlations between TUG measures and attentional performance (Spearman's r = .30-.54, p < .05-.01), as well as postural stability indices (e.g., unstable platform with eyes open: r = .38-.57, p < .05-.01). These findings demonstrate the feasibility and validity of using OpenTUG to automatically extract clinically relevant gait parameters. The observed associations between TUG performance, attentional measures, and postural stability highlight the potential of combining simple walking tasks with pose estimation as a rapid, cost-effective preliminary assessment of vestibular patients' difficulties.