Quantifying cortical lesions in multiple sclerosis MRI datasets using multi-contrast post-processing and deep...

Area of Science:

• Neurology
• Radiology
• Medical Imaging

Background:

• Multiple sclerosis (MS) is a chronic neurological disease impacting the central nervous system.
• Cortical lesions are key in MS, but difficult to visualize with standard MRI, limiting clinical trial analysis.
• Advanced MRI post-processing techniques, including AI, show promise for improved cortical lesion detection.

Purpose of the Study:

• To evaluate a unified framework combining and extending existing methods for enhanced cortical lesion detection in MS.
• To assess the feasibility of using synthetic contrasts and AI-based approaches on conventional MRI data.
• To reanalyze existing clinical trial data for mechanistic insights into MS development and treatment effects.

Main Methods:

• Utilized data from the large, multicenter, phase 3 ORATORIO trial.
• Evaluated fluid-attenuated inversion recovery squared (FLAIR²), T1/T2 ratio, and AI-derived double inversion recovery (AI-DIR).
• Introduced a novel multi-modal cortical lesion enhanced (MMCLE) contrast and employed transformer-based semantic segmentation for automated lesion detection.

Main Results:

• Detected an average of 14.8 ± 20.72 lesions per participant at baseline.
• Achieved an 86.0% true positive rate and 8.4% false positive rate with blinded MMCLE.
• Demonstrated high reproducibility across different field strengths and acquisition types (ICC 88.8-92.5%).

Conclusions:

• Confirmed that advanced MRI methods enable clear visualization and quantification of cortical lesions in MS.
• Validated that the simultaneous use of multiple contrasts, enhanced by deep learning, significantly improves lesion quantification.