Related Experiment Video
Updated: Jan 19, 2026
Comprehensive Autopsy Program for Individuals with Multiple Sclerosis
Published on: July 19, 2019
Predicting PET-derived demyelination from multimodal MRI using sketcher-refiner adversarial training for multiple
Wen Wei1, Emilie Poirion2, Benedetta Bodini3
1Université Côte d'Azur, Inria, Epione Project-Team, Sophia Antipolis, France; Inria, Aramis project-team, Paris, France; Institut du Cerveau et de la Moelle épinière, ICM, Inserm U 1127, CNRS UMR 7225, Sorbonne Université, Paris F-75013, France.
Abstract:
Multiple sclerosis (MS) is the most common demyelinating disease. In MS, demyelination occurs in the white matter of the brain and in the spinal cord. It is thus essential to measure the tissue myelin content to understand the physiopathology of MS, track progression and assess treatment efficacy. Positron emission tomography (PET) with [11C]PIB is a reliable method to measure myelin content in vivo. However, the availability of PET in clinical centers is limited. Moreover, it is expensive to acquire and invasive due to the injection of a radioactive tracer. By contrast, MR imaging is non-invasive, less expensive and widely available, but conventional MRI sequences cannot provide a direct and reliable measure of myelin. In this work, we therefore propose, to the best of our knowledge for the first time, a method to predict the PET-derived myelin content map from multimodal MRI. To that purpose, we introduce a new approach called Sketcher-Refiner generative adversarial networks (GANs) with specifically designed adversarial loss functions. The first network (Sketcher) generates global anatomical and physiological information. The second network (Refiner) refines and generates the tissue myelin content. A visual attention saliency map is also proposed to interpret the attention of neural networks. Our approach is shown to outperform the state-of-the-art methods in terms of image quality and myelin content prediction. Particularly, our prediction results show similar results to the PET-derived gold standard at both global and voxel-wise levels indicating the potential for clinical management of patients with MS.
Insights
This study introduces a novel method using generative adversarial networks (GANs) to predict myelin content maps from MRI scans for multiple sclerosis (MS) patients. This approach offers a non-invasive, cost-effective alternative to PET imaging for tracking MS progression.
Area of Science:
- Neuroimaging
- Medical Physics
- Artificial Intelligence
Background:
- Multiple sclerosis (MS) is a leading cause of demyelination in the central nervous system, necessitating accurate myelin content measurement for disease management.
- Positron emission tomography (PET) with [11C]PIB accurately measures in vivo myelin but faces limitations in clinical availability, cost, and invasiveness.
- Conventional magnetic resonance imaging (MRI) is widely accessible and non-invasive but lacks direct, reliable myelin quantification capabilities.
Purpose of the Study:
- To develop and validate a novel method for predicting in vivo myelin content maps from multimodal MRI data.
- To overcome the limitations of PET imaging in terms of accessibility, cost, and invasiveness for myelin quantification in MS.
- To establish a reliable, non-invasive imaging biomarker for MS assessment using advanced AI techniques.
Main Methods:
- Introduction of a novel generative adversarial network (GAN) framework named Sketcher-Refiner, comprising two networks for sequential myelin content prediction.
- Development of specialized adversarial loss functions tailored for myelin content estimation and image quality enhancement.
- Integration of a visual attention saliency map for interpreting the neural network's focus during myelin prediction.
Main Results:
- The proposed Sketcher-Refiner GAN approach significantly outperforms existing state-of-the-art methods in both image quality and myelin content prediction accuracy.
- Predicted myelin content maps demonstrated high concordance with the gold standard PET-derived measurements at both global and voxel-wise levels.
- The method shows potential for accurate, non-invasive assessment of myelin in the clinical management of multiple sclerosis.
Conclusions:
- The developed AI-driven MRI method provides a promising, non-invasive alternative for accurate myelin quantification in multiple sclerosis.
- This approach enhances the clinical utility of widely available MRI technology for tracking MS progression and treatment response.
- Further validation could integrate this technique into routine clinical practice for improved patient care in MS.
Related Concept Videos
09:41Comprehensive Autopsy Program for Individuals with Multiple Sclerosis
10:46MRI and PET in Mouse Models of Myocardial Infarction
11:35The Multiple Sclerosis Performance Test (MSPT): An iPad-Based Disability Assessment Tool
08:11Measuring Progressive Neurological Disability in a Mouse Model of Multiple Sclerosis
08:48Adapted Resistance Training Improves Strength in Eight Weeks in Individuals with Multiple Sclerosis
08:18A Protocol for the Use of Remotely-Supervised Transcranial Direct Current Stimulation (tDCS) in Multiple Sclerosis (MS)