Related Experiment Video
Updated: Sep 8, 2025
04:25
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
2.8K
Framework to Segment and Evaluate Multiple Sclerosis Lesion in MRI Slices Using VGG-UNet
Sujatha Krishnamoorthy1,2, Yaxi Zhang3, Seifedine Kadry4
1Zhejiang Bioinformatics International Science and Technology Cooperation Center, Wenzhou-Kean University, Wenzhou, Zhejiang Province, China.
Computational Intelligence and Neuroscience
|June 13, 2022
Summary
This study introduces a VGG-UNet convolutional neural network (CNN) for improved multiple sclerosis (MS) lesion detection in brain MRI scans. The developed method significantly enhances lesion segmentation accuracy for earlier diagnosis and treatment.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Healthcare
- Neuroscience and Neurology
Background:
- Multiple sclerosis (MS) is a debilitating autoimmune disease affecting the central nervous system (CNS).
- Early detection and precise lesion segmentation in brain Magnetic Resonance Imaging (MRI) are crucial for effective patient management and treatment.
- Existing segmentation methods often face challenges in accurately identifying and delineating MS lesions.
Purpose of the Study:
- To implement and evaluate a novel VGG-UNet convolutional neural network (CNN) segmentation scheme for extracting multiple sclerosis (MS) lesions from 2D brain MRI slices.
- To enhance the accuracy and efficiency of MS lesion detection compared to traditional segmentation techniques.
- To provide a robust tool for early diagnosis and monitoring of MS progression.
Main Methods:
- A VGG-UNet architecture was developed, utilizing a pre-trained VGG19 network as the encoder component for feature extraction.
- The proposed CNN segmentation scheme was trained and tested on a dataset of 600 2D brain MRI slices (512 × 512 × 3 pixels) from 30 patients.
- Performance was evaluated using metrics such as Jaccard index, Dice coefficient, and accuracy across axial, coronal, and sagittal planes of Flair MRI sequences.
Main Results:
- The VGG-UNet scheme demonstrated superior performance in MS lesion segmentation compared to traditional UNet, SegNet, VGG-UNet, and VGG-SegNet.
- Achieved high segmentation accuracy with Jaccard index exceeding 85%, Dice coefficient over 92%, and overall accuracy surpassing 98%.
- The model proved effective across multiple imaging planes (axial, coronal, sagittal) in Flair MRI modality.
Conclusions:
- The VGG-UNet CNN segmentation scheme offers a significant advancement in the automated detection and segmentation of MS lesions in brain MRI.
- This approach holds promise for improving the speed and reliability of MS diagnosis, potentially leading to better patient outcomes.
- Further research can explore the integration of this method into clinical workflows for real-time MS monitoring.
