Related Experiment Video
Updated: Jan 17, 2026
06:18
Author Spotlight: Segmentation and VR for Advanced Neurovascular Interventions
Published on: April 5, 2024
1.5K
MambaDiff: Mamba-Enhanced Diffusion Model for 3D Medical Image Segmentation.
Summary
This study introduces MambaDiff, a novel Mamba-enhanced diffusion model for 3D medical image segmentation. MambaDiff improves segmentation accuracy by enhancing feature extraction and preserving structural consistency, outperforming existing methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Accurate 3D medical image segmentation is vital for clinical diagnosis and treatment planning.
- Diffusion models show potential in medical image segmentation but struggle with fine-grained feature preservation and structural consistency due to weak conditional guidance and limited feature extraction.
- Existing methods often fail to capture intricate details and maintain spatial integrity in segmentation tasks.
Purpose of the Study:
- To develop an advanced diffusion model that overcomes the limitations of current approaches in 3D medical image segmentation.
- To enhance the accuracy and structural consistency of medical image segmentation by improving feature extraction and conditional information integration.
- To introduce a novel Mamba-enhanced diffusion model, MambaDiff, for superior performance in 3D medical image segmentation.
Main Methods:
- Proposed MambaDiff, a Mamba-enhanced diffusion model for 3D medical image segmentation.
- Employed an encoder to extract multilevel semantic features from original images.
- Integrated features using a Semantic Hierarchical Embedding (SHE) mechanism and incorporated a Global-Slice Perception Mamba (GSPM) layer for enhanced spatial reasoning and feature extraction.
Main Results:
- MambaDiff demonstrated competitive performance against state-of-the-art methods on four public medical image segmentation datasets (BraTS 2021, BraTS 2024, LiTS, MSD Hippocampus).
- The proposed model achieved superior segmentation accuracy and structural consistency.
- MambaDiff achieved these results with significantly fewer parameters compared to existing approaches.
Conclusions:
- MambaDiff effectively addresses the limitations of standard diffusion models in 3D medical image segmentation.
- The integration of Mamba architecture and the proposed SHE and GSPM mechanisms significantly enhances feature extraction and preserves structural integrity.
- The model offers a more parameter-efficient and accurate solution for critical medical image segmentation tasks.
