Brain age in multiple sclerosis: a study with deep learning and traditional machine learning.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Neurodegenerative Diseases

Background:

• Accelerated brain aging in multiple sclerosis (MS) correlates with disability.
• Artificial intelligence (AI) offers tools for quantifying neurodegeneration.
• Comparative data on traditional machine learning (ML) versus deep learning (DL) for brain age in MS is lacking.

Purpose of the Study:

• To validate a deep learning (DL) brain age model in multiple sclerosis (MS).
• To compare DL and traditional ML models for estimating brain age in MS.
• To assess the association of brain age with clinical outcomes and scanner variability.

Main Methods:

• Retrospective analysis of 4584 MRI scans from 1516 MS patients.
• Comparison of a DL simple fully convolutional network with a traditional ML model for brain age estimation.
• Analysis of clinical and MRI data from a longitudinal cohort using a uniform post-processing pipeline.

Main Results:

• DL brain age estimates showed a stronger correlation with chronological age (r=0.90) than traditional ML (r=0.75).
• Increased brain age, estimated by both methods, significantly correlated with disability (EDSS scores) and disease duration.
• DL models demonstrated less scanner variability compared to traditional ML.

Conclusions:

• Deep learning-derived brain age is a valid measure in MS, strongly associated with clinical disability.
• DL brain age estimation performs comparably to traditional ML but may offer improved robustness against scanner differences.
• AI-driven brain age assessment holds promise for monitoring neurodegeneration in MS.