Related Experiment Video
Updated: Jun 12, 2026
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
47.9K
Contrast-Enhancing Lesion Segmentation in Multiple Sclerosis: A Deep Learning Approach Validated in a Multicentric
Martina Greselin1,2,3, Po-Jui Lu1,2,3, Lester Melie-Garcia1,2,3
1Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, Faculty of Medicine, University Hospital Basel, University of Basel, 4123 Basel, Switzerland.
Bioengineering (Basel, Switzerland)
|August 29, 2024
Summary
This study developed a deep learning model for automatic detection of contrast-enhancing lesions (CELs) in multiple sclerosis (MS) using MRI scans. The AI model shows promise for improving diagnostic accuracy and reducing variability in clinical practice.
Area of Science:
- Neurology
- Medical Imaging
- Artificial Intelligence
Background:
- Contrast-enhancing lesions (CELs) are crucial for diagnosing and monitoring multiple sclerosis (MS).
- Manual detection of CELs is time-consuming and prone to significant variability among clinicians.
- Automated methods for CEL detection are limited but essential for improving clinical workflow.
Purpose of the Study:
- To develop and evaluate a deep learning model for automatic detection and segmentation of CELs in brain MRI scans of MS patients.
- To assess the performance of the 3D UNet-based model in a clinical setting.
Main Methods:
- A 3D UNet deep learning architecture was employed for CEL detection and segmentation.
- The model was trained on a dataset of 372 MRI scans (T1-weighted pre/post-gadolinium, FLAIR) from 280 MS patients.
- A weighted loss function and lesion mask sampling strategy were used to address dataset imbalance and improve accuracy.
Main Results:
- The deep learning model achieved a Dice Score Coefficient of 0.76.
- The model demonstrated high accuracy with a True Positive Rate of 0.93 and a low False Positive Rate of 0.02.
- The developed model shows significant potential for accurate and consistent CEL detection.
Conclusions:
- The developed deep learning model effectively detects and segments CELs in MS patients' MRI scans.
- This automated approach has the potential to enhance clinical decision-making and reduce diagnostic variability.
- The model could serve as a valuable tool for supporting radiologists and neurologists in MS management.
