Related Experiment Video
Updated: Jul 10, 2026
10:26
A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules
Published on: May 19, 2023
Efficient low-dose CT image enhancement using MobileMamba-UNet with wavelet-enhanced long-range modeling.
Jianfang Li1,2, Haiyan Liu3, Xiaoli Wang4
1School of Information Engineering, Changsha Medical University, Changsha, China.
Journal of Applied Clinical Medical Physics
|July 8, 2026
Summary
MobileMamba-UNet offers efficient low-dose computed tomography (LDCT) reconstruction by integrating global and local details. This deep learning model achieves superior image quality with reduced memory and faster speeds, enhancing practical applicability.
Area of Science:
- Medical Imaging
- Deep Learning
- Computational Efficiency
Background:
- Deep learning dominates low-dose computed tomography (LDCT) image reconstruction.
- Existing methods face challenges in balancing structural detail preservation with computational efficiency, especially for long-range dependencies.
Purpose of the Study:
- To develop a lightweight LDCT reconstruction framework.
- The framework aims to capture global context and local details efficiently.
- It prioritizes low memory consumption and fast inference speed.
Main Methods:
- Proposed MobileMamba-UNet, a hybrid neural network combining MobileMamba backbone and U-Net architecture.
- Incorporated Wavelet Transform Enhanced Mamba for high-frequency structures.
- Utilized multi-receptive field feature interaction for local textures and long-range dependencies.
- Ensured linear computational complexity for large-scale LDCT tasks.
Main Results:
- MobileMamba-UNet outperformed existing CNN- and Transformer-based methods on Mayo-2016 and Mayo-2020 LDCT datasets.
- Achieved superior image quality compared to existing approaches.
- Significantly reduced memory usage and inference latency.
Conclusions:
- MobileMamba-UNet presents a promising solution for LDCT image reconstruction.
- The model effectively balances reconstruction performance with computational efficiency.
- Demonstrates practical applicability for real-world LDCT imaging.
