Therapy effect on AI-derived thalamic atrophy using clinical routine MRI protocol: A longitudinal, multi-center,...

Area of Science:

• Neuroimaging
• Artificial Intelligence in Medicine
• Neurology

Background:

• Assessing the impact of disease-modifying therapies (DMTs) on brain atrophy in persons with multiple sclerosis (pwMS) typically requires standardized clinical trials or single-center data.
• There is a need to evaluate DMT efficacy using real-world, unstandardized neuroimaging data.

Purpose of the Study:

• To determine the effect of DMTs on lateral ventricular volume (LVV) and thalamic volume (TV) changes in pwMS.
• To utilize artificial intelligence (AI)-based volumetric analysis on routine, unstandardized T2-FLAIR MRI scans for this evaluation.

Main Methods:

• The DeepGRAI registry enrolled 1002 relapsing-remitting pwMS across 30 US sites in a longitudinal, observational study.
• Routine clinical MRI scans (T2-FLAIR) were collected at baseline and ~2.6 years follow-up, acquired on unharmonized 1.5T or 3T scanners.
• AI tools (DeepGRAI for TV, NeuroSTREAM for LVV) were used for volumetric analysis, with propensity matching applied.

Main Results:

• Untreated pwMS showed significantly greater thalamic volume (TV) loss compared to treated pwMS (-1.2% vs. -0.3%, p=0.044).
• High-efficacy DMTs resulted in a two-fold lower lateral ventricular volume (LVV) change compared to moderate-efficacy DMTs (3.5% vs. 7.0%, p=0.001).
• Patients who stopped DMT experienced greater TV and LVV changes than those who continued treatment.

Conclusions:

• AI-driven volumetric analysis of LVV and TV on routine T2-FLAIR scans can detect short-term, treatment-induced neurodegenerative changes.
• This approach is effective in a real-world, unstandardized, multicenter clinical setting, demonstrating the utility of AI in assessing DMT effects.