3D Classification of Paramagnetic Rim Lesions in Multiple Sclerosis via Asymmetric QSM-FLAIR Modeling
Researchers have developed a novel 3D multimodal deep learning framework to classify paramagnetic rim lesions (Rim+) in multiple sclerosis (MS) patients using Quantitative Susceptibility Mapping (QSM) and FLAIR MRI. This approach models modality asymmetry by using QSM as the primary signal, conditioned by FLAIR structural context. The method utilizes self-supervised multimodal pretraining followed by supervised fine-tuning to enhance robustness with limited data. Evaluated on a cohort of 88 MS patients, the framework demonstrated improved performance over previous architectures for automated identification of chronic active lesions. AI
IMPACT This research could lead to more accurate and efficient diagnosis of chronic active inflammation in multiple sclerosis patients.