# Using AI-Based Gait Analysis to Establish a 5-Meter Walk Time Cutoff for Discriminating Alzheimer's Disease

## TL;DR

This study uses AI-powered gait analysis to identify a 5-meter walk time cutoff that can help detect Alzheimer's disease in elderly individuals.

## Contribution

The study introduces a practical AI-based gait analysis method with specific cutoff values for early Alzheimer's detection.

## Key findings

- Alzheimer's patients showed significantly slower gait speed (12.12 vs. 3.98 sec/5 m) compared to healthy controls.
- A gait speed cutoff of 5.85 sec/5 m achieved 96.7% sensitivity and 96.5% specificity in detecting Alzheimer's.
- Gait rhythm and left-right asymmetry also showed effective cutoff values for dementia discrimination.

## Abstract

Introduction

A decline in gait function has been reported to occur early in Alzheimer's disease (AD), a major cause of dementia, suggesting that gait analysis may be a useful tool for dementia screening. However, a simple and practical analysis method or clear cutoff values have yet to be established. This study aimed to analyze gait function using the AI-powered smartphone application "Toruto," identify gait indicators characteristic of elderly patients with AD, and propose effective cutoff values for dementia discrimination.

Methods

A total of 147 participants were included in the study: 86 healthy elderly individuals and 61 elderly patients with AD (102 female patients and 45 male patients). Exclusion criteria included the use of a cane, the presence of pain during walking, or the need for walking assistance. Gait function at a normal walking speed was analyzed by the AI of "Toruto" to assess speed, rhythm, and left-right asymmetry. An unpaired t-test was used to compare the two groups, and cutoff values for dementia discrimination were calculated using receiver operating characteristic (ROC) curve analysis. The significance level was set at p < 0.05.

Results

The AD group showed a significant decline in gait speed (12.12 versus 3.98 sec/5 m), rhythm (10.92 versus 6.51), and left-right asymmetry (6.34 versus 2.15) compared with the healthy control group (p < 0.05). ROC curve analysis revealed that using a gait speed cutoff value of 5.85 sec/5 m yielded a sensitivity of 96.7% and a specificity of 96.5%. A rhythm cutoff of 7.06 (sensitivity: 80.3%, specificity: 65.1%) and a left-right asymmetry cutoff of 2.25 (sensitivity: 75.4%, specificity: 67.4%) were also effective discriminative indicators.

Conclusion

Gait analysis using the AI-powered smartphone application "Toruto" is effective for distinguishing Alzheimer's disease. This study demonstrated that cutoff values, such as 5.85 seconds for a 5-meter walk, serve as practical screening indicators for dementia. As an objective biomarker reflecting cognitive decline, gait function is expected to be useful for the early detection of Alzheimer's disease in both community and clinical settings.