Predicting minimum toe clearance in older adults on different surfaces from level ground data: A machine learning...

Area of Science:

• Biomechanics
• Gerontology
• Gait Analysis

Background:

• Trips are common on varied surfaces, yet minimum toe clearance (MTC) is typically measured on level ground.
• Existing research shows MTC differences between level and non-level walking.
• Predictive capability of lab-based MTC for real-world, challenging terrains remains unclear.

Purpose of the Study:

• To determine if level-ground MTC predicts MTC on carpet and obstacle surfaces in older adults.
• To assess the transferability of laboratory gait data to complex walking conditions.

Main Methods:

• Twenty-one older adults performed gait trials on level, carpet, and 5mm/13mm obstacle surfaces.
• Fifteen level-ground gait variables were used as predictors.
• Nested cross-validation and principal component analysis (PCA) were employed for model development.

Main Results:

• A k-nearest neighbors with PCA (kNN-PCA) model accurately predicted toe-tip MTC (CCC 0.79-0.90, MAE 2.30-3.84mm).
• Prediction accuracy decreased with increasing obstacle height.
• Right-foot MTC was generally more predictable than left-foot MTC.

Conclusions:

• Level-ground gait measurements can reliably estimate MTC on carpet and obstacle surfaces in older adults.
• Laboratory MTC assessments are a practical tool for estimating toe clearance in challenging environments.
• Findings support using lab data for fall-risk assessment and rehabilitation planning.