A Multiple Regression Approach to Normalization of Spatiotemporal Gait Features

Area of Science:

• Biomechanics
• Neurology
• Data Science

Background:

• Gait analysis is crucial for understanding human movement.
• Intersubject variability in physical characteristics complicates gait data interpretation.
• Existing normalization methods may not fully account for these variations.

Purpose of the Study:

• To develop and evaluate a multiple regression normalization approach for spatiotemporal gait data.
• To account for intersubject variations in walking speed and physical properties.
• To enhance the detection of pathological gait patterns, particularly in Parkinson's disease (PD).

Main Methods:

• Collected spatiotemporal gait data from diverse age groups and PD patients.
• Applied standard dimensionless equations, detrending, and a novel multiple regression normalization.
• Compared the effectiveness of different normalization techniques in reducing correlations and identifying group differences.

Main Results:

• Multiple regression normalization significantly reduced correlations between gait parameters and physical properties compared to other methods.
• This approach enhanced the identification of significant differences in multiple gait parameters for PD patients.
• Weak to moderate correlations were observed after dimensionless and detrending normalization, which were reduced to weak values using multiple regression.

Conclusions:

• The proposed multiple regression normalization effectively reduces intersubject variability in gait data.
• This method improves the sensitivity of gait analysis for detecting pathological gait patterns.
• The approach shows promise for applications in machine learning, gait classification, and clinical gait evaluation.