Dementia Prediction Using Gait Analysis and Machine Learning

Dementia is a progressive neurodegenerative disorder affecting millions of people worldwide. Early prediction of dementia, especially during the mild cognitive impairment (MCI) stage, is crucial for timely intervention and management. Gait analysis provides indicators of cognitive decline and help identify individuals at risk of progression. This study aims to extract gait features from video recordings and apply machine learning models (Support Vector Machine (SVM), XGBoost, and Logistic Regression (LR)) to predict conversion from MCI to dementia. The dataset consists of videos from 62 individuals with MCI, of whom 31 later converted to dementia at 2-year follow-up. Participants performed the Timed Up and Go (TUG) test under single-task and dual-task (TUGdt) conditions, including animal naming (TUGdt-NA) and reciting months in reverse order (TUGdt-MB). The results showed that SVM achieved the highest performance with an accuracy of 70% and F1 score of 69%. These findings show that gait-based machine learning models, particularly SVM, show promise for early prediction of dementia conversion in individuals with MCI.