MS-DASPNet: Multiple Sclerosis lesion segmentation from brain MRI using dual attention and spatial pyramid pooling
Shikha Jain1, Navin Rajpal1, Pramod Kumar Soni2
1University School of Information, Communication & Technology, Guru Gobind Singh Indraprastha University, New Delhi, India.
Frontiers in Computational Neuroscience
|February 16, 2026
Summary
Accurate segmentation of multiple sclerosis (MS) lesions in brain MRI is crucial for patient care. A new Dual Attention Guided Deep Neural Network (MS-DASPNet) demonstrates superior performance in detecting these challenging lesions.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neurology
Background:
- Accurate detection and segmentation of multiple sclerosis (MS) lesions in brain Magnetic Resonance Imaging (MRI) is challenging due to lesion characteristics like small size, irregular shape, and variability across imaging modalities.
- Precise MS lesion segmentation is vital for early diagnosis, monitoring disease progression, and guiding treatment planning.
Purpose of the Study:
- To introduce MS-DASPNet, a novel Dual Attention Guided Deep Neural Network designed to overcome challenges in MS lesion detection, including small size, low contrast, and heterogeneous appearance.
- To evaluate the effectiveness of MS-DASPNet against state-of-the-art methods on multiple public datasets.
Main Methods:
- MS-DASPNet utilizes a VGG-16-based encoder and an Atrous Spatial Pyramid Pooling (ASPP) bottleneck for multi-scale context learning.
- Dual attention modules are incorporated into skip connections to refine spatial details and enhance channel-wise feature representation.
- The network was evaluated on four public datasets: ISBI-2015, Mendeley, MICCAI-2016, and MICCAI-2021.
Main Results:
- MS-DASPNet achieved superior Precision, Dice, Sensitivity, and Jaccard scores compared to existing state-of-the-art methods across the evaluated datasets.
- The model attained a Dice score of 0.8736 on the MICCAI-2016 dataset and 0.8706 on the MICCAI-2021 dataset.
- These results indicate enhanced performance over current segmentation techniques.
Conclusions:
- MS-DASPNet demonstrates robustness and effectiveness in the accurate segmentation of multiple sclerosis lesions from brain MRI.
- The proposed Dual Attention Guided Deep Neural Network offers a promising advancement for clinical applications requiring precise MS lesion segmentation.
