Integrating Mamba Sequence Model and Hierarchical Upsampling Network for Accurate Semantic Segmentation of Multiple Sclerosis Lesion

Integrating components from convolutional neural networks and state space models in medical image segmentation presents a compelling approach to enhance accuracy and efficiency. We introduce Mamba-HUNet, a novel architecture tailored for robust and efficient segmentation tasks. Leveraging strengths from Mamba-UNet and the lighter version of Hierarchical Upsampling Network (HUNet), Mamba-HUNet combines convolutional neural networks’ local feature extraction power with state space models’ long-range dependency modeling capabilities. The lighter HUNet was integrated into Mamba-HUNet, maintaining performance parity while improving computational efficiency. Mamba-HUNet improves existing semantic segmentation methods by efficiently capturing localized fine-grained features and long-range dependencies in medical images through Visual State Space blocks and patch merging layers. This results in enhanced segmentation accuracy and robustness. Experimental results on publicly available Magnetic Resonance Imaging scans, notably in Multiple Sclerosis lesion segmentation, demonstrate Mamba-HUNet’s superior performance across diverse segmentation tasks, highlighting its potential in improving clinical decision-making processes.