Tensor electrical impedance myography identifies clinically relevant features in amyotrophic lateral sclerosis

Area of Science:

• Biomedical Engineering
• Neurology
• Data Science

Background:

• Electrical impedance myography (EIM) is a promising biomarker for amyotrophic lateral sclerosis (ALS).
• EIM uses multiple frequencies and electrode configurations to assess muscle properties.
• High-dimensional EIM data presents challenges for identifying clinically relevant features.

Purpose of the Study:

• To evaluate non-negative tensor factorization (NTF) as a framework for analyzing high-dimensional EIM data.
• To determine if NTF can identify clinically relevant features in EIM datasets for ALS.
• To compare the performance of tensor EIM with raw data and feature selection methods.

Main Methods:

• EIM data, including resistivity and reactivity at 14 frequencies across three electrode configurations, were collected from healthy individuals and ALS patients.
• Non-negative tensor factorization (NTF) was applied for dimensionality reduction, creating 'tensor EIM'.
• Statistical significance tests, symptom correlation analyses, and classification approaches were employed.

Main Results:

• Tensor EIM significantly differentiated between healthy and ALS patients (p<0.001, AUROC=0.78).
• Tensor EIM also distinguished between mild and severe ALS disease states (p<0.001, AUROC=0.75).
• A significant correlation was found between tensor EIM metrics and ALS symptoms (ρ=0.7, p<0.001), with a trend towards rightward spectral shifts in diseased muscle, consistent with atrophy.

Conclusions:

• Tensor EIM provides clinically relevant metrics for identifying ALS-related muscle disease.
• This method utilizes the entire spectral dataset, reducing the risk of overfitting.
• The approach identifies disease-specific spectral shapes, enabling deeper clinical interpretation of muscle changes in ALS.