Overview
Researchers have developed a generative AI framework called Multimodal Cortical Lesion Enhancement (MMCLE) that can detect gray matter lesions in multiple sclerosis patients that are invisible to conventional MRI analysis. The system works on legacy scans, meaning clinics can re-analyze existing patient data without acquiring new imaging hardware.
The Problem
Gray matter (cortical) lesions in MS are strongly correlated with disability progression and cognitive decline, but they are notoriously difficult to detect on standard MRI sequences. Many patients whose disease appears stable on conventional imaging may actually have significant cortical pathology driving their symptoms.
How MMCLE Works
The AI framework uses a generative model trained on paired data from specialized research MRI sequences (like double inversion recovery and phase-sensitive inversion recovery) and standard clinical sequences. Once trained, it can hallucinate the enhanced contrast of specialized sequences from standard scans, revealing lesions that would otherwise require expensive, time-consuming research protocols.
Impact
Across study cohorts, MMCLE identified over 11,000 cortical lesions that had been missed by conventional analysis. This could dramatically improve disease monitoring, treatment decisions, and clinical trial endpoint sensitivity — all without requiring patients to undergo additional imaging.
