Walking into the future – Enhancing cognitive diagnosis in vestibular patients using walking task & deep pose estimation.

Purpose Vestibular disorders impact body orientation, navigation, with varying degrees of affective and cognitive symptoms, reducing patients' quality of life. Previous studies have assessed these effects, usually with extensive and time-consuming physiological and neuropsychological assessment. This study aims to explore the potential of a simple walking task as an indicator of cognitive issues in vestibular disorder patients. Methods Sixty-five patients with mixed vestibular disorders, undergoing vestibular physiotherapy, completed a Timed-Up & Go (TUG) task alongside neuropsychological evaluations. Attention was measured using tasks from the Test for Attentional Performance (TAP), while quality of life was assessed through questionnaires. Video recordings of the TUG task were analyzed using a 2D human pose estimation method (OpenTUG) for step count and time taken. Same variables were manually recorded by experimenters for comparison. A control group of thirty average-age matched healthy controls also performed the Timed-Up & Go (TUG) task and the neuropsychological evaluation. Results Patients took significantly more time and performed more steps in the TUG task than healthy controls at normal and slow speeds. The observed results were similar between the automatic steps/time derived from OpenTUG relative to steps/time analyses by experimenters, with highly significant correlations. Significant medium to high correlations were detected between cognitive reaction time of alertness measures and the visual subtask of the divided attention TAP task. Significant medium correlations were also present between the DHI and the number of steps/time taken during the TUG task. Conclusions These results demonstrated that it is possible to automatically extract pathology-relevant gait parameters using deep pose estimation on 2-D video-recordings of patients with vestibular disorders. The observed relationship between the TUG task and neuropsychological subjective and objective measures demonstrates a need to further explore the feasibility of using a simple walking task combined with pose estimation. It could provide a first line assessment before carrying out extensive neuropsychological evaluation.