An interpretable machine learning model to explain the interplay between brain lesions and cortical atrophy in...

Area of Science:

• Neuroimaging
• Neurology
• Medical Physics

Background:

• Multiple Sclerosis (MS) is a leading cause of neurological disability in young adults.
• Cortical grey matter atrophy correlates with MS progression, but its underlying causes, especially lesion relationships, are unclear.

Purpose of the Study:

• To investigate the relationship between cortical atrophy and different lesion types in MS using Ultra-High Field (UHF) 7 Tesla MRI.
• To explore the interplay of cortical lesions, white matter (WM) lesions, and lesions with susceptibility rims in MS pathology.

Main Methods:

• Utilized UHF 7 Tesla MRI for detailed lesion characterization.
• Applied a machine learning (ML) framework with explainability to predict cortical atrophy from MRI-derived lesion features.
• Analyzed lesion types including intracortical lesions and those with susceptibility rims.

Main Results:

• Successfully predicted cortical atrophy in MS patients using a limited set of lesion-related features from 7 T MRI.
• Identified WM lesion volume (excluding rims), patient age, WM lesion count (excluding rims), and intracortical lesion volume as key predictors.
• Observed contrasting roles for WM and cortical lesion load, with small WM lesions and larger cortical lesions being significant.

Conclusions:

• Cortical atrophy in MS is significantly influenced by white matter lesion-induced disconnection and axonal degeneration, alongside local cortical demyelination.
• UHF MRI is crucial for accurate assessment of intracortical and rim lesions, highlighting their distinct contributions to MS severity and progression.
• Advanced imaging and analytical approaches are essential for understanding the complex etiopathogenesis of MS lesions.