Related Experiment Video
Updated: Jun 8, 2026
Comprehensive Autopsy Program for Individuals with Multiple Sclerosis
Published on: July 19, 2019
Diagnosis of Multiple Sclerosis Using Multimodal Deep Learning Integrating Lesion and Normal-Appearing White Matter:
Jiajian Ma1, Valentin Stepanov2,3,4, Wushuang Rui5
1Center for Data Science, New York University, New York, NY, USA.
Abstract:
Current multiple sclerosis (MS) diagnosis relies primarily on focal white matter lesions (WMLs), which are frequently mimicked by other conditions. Normal-appearing white matter (NAWM) harbours complementary pathological information but remains clinically underutilised because NAWM alterations are macroscopically occult on routine scans and require non-routine quantitative imaging to visualise. Here, we show that NAWM-related diagnostic information can be recovered from routine structural MRI using a cross-modal deep-learning model. We developed DeepMS, a model co-trained on diffusion and structural MRI that operates solely on structural MRI at deployment. DeepMS achieved ROC-AUCs of 0.968 internally (n=837) and 0.940-0.974 across two international external cohorts (n=293 and n=1,756). In a multi-reader study, DeepMS outperformed the 2024 McDonald criteria imaging biomarkers. DeepMS retained robust performance after digital lesion removal and exhibited NAWM-dominant activation maps. Combined with established imaging biomarkers, DeepMS improved sensitivity (92.1% vs 74.8%) while maintaining high specificity (95.6% vs 92.3%) compared with corresponding biomarker composite based on the 2024 McDonald criteria. By decoding latent NAWM signals from routine scans and integrating them with WML features, this framework can potentially advance MS diagnosis beyond the current lesion-centric paradigm.
Insights
Deep learning model DeepMS accurately diagnoses multiple sclerosis (MS) using routine MRI scans by analyzing normal-appearing white matter (NAWM) and white matter lesions (WMLs). This approach improves diagnostic specificity and sensitivity, aiding in cases with ambiguous abnormalities.
Area of Science:
- Artificial Intelligence in Medicine
- Neuroimaging Analysis
- Machine Learning for Diagnostics
Background:
- Current multiple sclerosis (MS) diagnosis relies on white matter lesions (WMLs), which lack specificity.
- Normal-appearing white matter (NAWM) abnormalities offer complementary diagnostic information.
- Establishing a reproducible NAWM signature for MS diagnosis is challenging with routine MRI.
Purpose of the Study:
- To develop and validate DeepMS, a deep learning model for MS diagnosis.
- To integrate both WML and NAWM features from routine MRI scans.
- To improve diagnostic accuracy beyond current criteria.
Main Methods:
- Developed DeepMS using quantitative diffusion MRI (dMRI) and structural MRI (sMRI) data from 8,450 scans.
- Validated DeepMS on internal and two independent external cohorts (totaling 3,076 patients).
- Compared DeepMS performance against 2024 McDonald criteria biomarkers and performed lesion-masking experiments.
Main Results:
- DeepMS achieved high AUCs across all cohorts (internal: 0.968, Krakow: 0.940, public external: 0.974).
- DeepMS demonstrated superior specificity and sensitivity compared to established MS biomarkers (DIS, CVS).
- The model retained diagnostic capability even after masking WMLs, highlighting the utility of NAWM features.
Conclusions:
- Deep learning models can effectively utilize NAWM information from routine sMRI for MS diagnosis.
- Integrating NAWM features enhances diagnostic capabilities, particularly for patients with ambiguous white matter abnormalities.
- DeepMS shows potential to improve the accuracy and efficiency of MS diagnosis.
