Multiple Sclerosis Classification Using the Local Divergence Exponent: Parameters Selection for State-Space...

Area of Science:

• Biomedical Engineering
• Neurology
• Rehabilitation Science

Background:

• Walking stability is impaired in people with multiple sclerosis (pwMS).
• Previous studies using the local divergence exponent (LDE) to assess walking stability in pwMS show varied results due to differing calculation methods.
• Standardizing LDE calculation is crucial for reliable comparisons and clinical application.

Purpose of the Study:

• To investigate how different state space reconstruction parameters for LDE calculation impact the classification accuracy of pwMS.
• To identify optimal parameters for LDE calculation to serve as a reliable mobility biomarker in multiple sclerosis.

Main Methods:

• 55 pwMS and 23 controls underwent a 5-minute walking test.
• LDE was calculated using three parameter sets (trial-specific, median, fixed d=5/τ=10) and various sensor data (vertical, mediolateral, anteroposterior accelerations, norm, 3D) from sternum and lumbar sensors.
• Quadratic Discriminant Analysis (QDA) was used to compare classification accuracy across different LDE calculation methods.

Main Results:

• The highest classification accuracy (84%) was achieved using LDE from sternum-mounted norm acceleration data with fixed parameters (d=5, τ=10), incorporating walking speed as a covariate.
• LDE calculations using lumbar sensors yielded lower classification accuracy compared to sternum sensors.

Conclusions:

• Fixed parameters (d=5, τ=10) for LDE calculation using sternum norm acceleration data provide the best classification of pwMS.
• Standardizing LDE calculation parameters simplifies its implementation as a mobility biomarker for multiple sclerosis.
• This study provides evidence supporting a consensus for LDE calculation methodology in MS research and clinical practice.