Differentiation between multiple sclerosis and neuromyelitis optic spectrum disorders with multilevel fMRI features:...

Area of Science:

• Neuroimaging
• Machine Learning
• Neurology

Background:

• Conventional statistical analysis of resting-state functional MRI (rs-fMRI) shows limitations in differentiating multiple sclerosis (MS) and neuromyelitis optic spectrum disorders (NMOSD).
• Improved diagnostic tools are crucial for accurate patient management in these neurological conditions.

Purpose of the Study:

• To develop and evaluate machine learning models for distinguishing MS from NMOSD using multilevel rs-fMRI metrics.
• To assess the added value of gray matter volume (GMV) data in classification.

Main Methods:

• Extracted multilevel functional metrics (RSFC, ALFF, ReHo) from 116 brain regions using the anatomical automatic labeling atlas.
• Developed and compared Support Vector Machine (SVM) and Logistic Regression (LR) classifiers integrating these features.
• Evaluated model performance on independent testing cohorts.

Main Results:

• Integrated models combining RSFC, ALFF, and ReHo achieved higher classification performance (SVM AUC=0.857, LR AUC=0.929) than models using individual features.
• Inclusion of GMV data did not significantly enhance classification accuracy.
• Optimal performance was consistent across different brain templates.

Conclusions:

• Multilevel rs-fMRI features (RSFC, ALFF, ReHo) provide a robust and effective method for differentiating MS and NMOSD patients.
• This approach offers a promising avenue for improving diagnostic efficacy in neuroimmunological disorders.