Dementia Prediction Using Gait Analysis and Machine Learning.

Area of Science:

• Neuroscience
• Gerontology
• Computer Science

Background:

• Dementia is a progressive neurodegenerative disorder impacting millions globally.
• Early prediction, particularly from mild cognitive impairment (MCI), is vital for intervention.
• Gait analysis offers potential biomarkers for cognitive decline and dementia risk.

Purpose of the Study:

• To extract gait features from video recordings.
• To apply machine learning models (SVM, XGBoost, LR) for predicting MCI to dementia conversion.
• To evaluate the efficacy of gait analysis in early dementia detection.

Main Methods:

• Utilized video recordings of 62 individuals with MCI, with 31 converting to dementia within 2 years.
• Employed the Timed Up and Go (TUG) test under single-task and dual-task (TUGdt-NA, TUGdt-MB) conditions.
• Extracted gait features and applied machine learning classifiers including Support Vector Machine (SVM), XGBoost, and Logistic Regression (LR).

Main Results:

• The Support Vector Machine (SVM) model demonstrated the highest predictive performance.
• SVM achieved an accuracy of 70% and an F1 score of 69% in predicting dementia conversion.
• Gait features extracted during TUG and TUGdt tasks were significant predictors.

Conclusions:

• Gait-based machine learning models show significant potential for the early prediction of dementia conversion in MCI patients.
• The SVM model, utilizing gait analysis, is a promising tool for identifying individuals at high risk of progressing to dementia.
• This approach supports the development of non-invasive methods for dementia risk assessment.