Functional data analysis for gait curves study in Parkinson's disease

Area of Science:

• Biomedical Engineering
• Statistics
• Neurology

Background:

• Accurate gait analysis is crucial for diagnosing Parkinson's disease (PD) and evaluating treatment efficacy.
• Traditional multivariate statistical methods are insufficient for analyzing high-dimensional, continuous gait curve data.
• Functional Data Analysis (FDA) provides advanced techniques for handling such complex data.

Purpose of the Study:

• To introduce an FDA approach for assessing gait curve reliability in PD patients.
• To enable intra-individual curve comparisons for evaluating therapeutic effects.
• To develop objective criteria for measuring gait improvements and comparing treatments.

Main Methods:

• Discretized gait curves were interpolated using cubic B-splines for continuity.
• Cluster analysis on smoothed curves identified reliable and representative gait trials.
• Functional Principal Component Analysis (FPCA) and Linear Discriminant Analysis (LDA) were used for curve comparison.

Main Results:

• The FDA approach successfully assessed gait curve reliability for individual patients.
• FPCA and LDA effectively highlighted differences between gait curves under various therapeutic conditions.
• Objective criteria were established for measuring gait improvements in 12 PD patients across 4 treatments.

Conclusions:

• FDA provides a robust framework for analyzing complex gait data in Parkinson's disease.
• The proposed methods enhance the objectivity of gait analysis and treatment evaluation.
• This FDA methodology holds potential for application in other medical fields with continuous curve data.