Machine learning classifier to identify clinical and radiological features relevant to disability progression in...

Area of Science:

• Neuroimaging
• Machine Learning
• Neurology

Background:

• Multiple Sclerosis (MS) is a chronic neurological disease characterized by unpredictable disability progression.
• Accurate prediction of MS disability progression is crucial for personalized treatment strategies.
• Current prediction methods often lack precision, highlighting the need for advanced analytical approaches.

Purpose of the Study:

• To assess the predictive accuracy of machine learning (ML) classification for disability progression in MS.
• To identify key neuroimaging and clinical features that contribute to predicting MS disability.
• To compare the performance of ML models based on radiological versus clinical data.

Main Methods:

• Analysis of structural brain MRI scans from 163 MS patients over 2-6 years.
• Calculation of T2-weighted lesion load (T2LL), gray matter volumes, and white matter integrity.
• Application of 1000 supervised ML classifiers using baseline imaging, clinical data, and follow-up disability status (Expanded Disability Status Scale - EDSS).

Main Results:

• 36% of participants experienced disability progression.
• The top-performing ML classifier achieved an accuracy of 0.79 and an Area Under the Curve (AUC) of 0.81.
• Key predictors included T2LL, thalamic volume, baseline EDSS, and therapy; radiological features outperformed clinical ones.

Conclusions:

• Machine learning models can effectively predict disability progression in MS.
• Neuroradiological features are primary drivers for accurate MS disability prediction using ML.
• This data-driven approach offers a promising tool for forecasting MS disease trajectories.