JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

A mixed Mamba U-net for prostate segmentation in MR images

  • 0Department of Urology, Hunan Provincial People's Hospital, The First Affiliated Hospital of Hunan Normal University, Changsha, 410005, People's Republic of China.
Scientific Reports +

|

August 28, 2024

|

August 28, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

This study introduces MM-UNet, a novel deep learning model for accurate prostate cancer segmentation in MRI scans. The method enhances diagnostic capabilities by improving segmentation performance on medical images.

Area Of Science

  • Medical Imaging Analysis
  • Deep Learning for Medical Diagnosis
  • Computational Anatomy

Background

  • Accurate prostate cancer diagnosis relies on precise segmentation of prostate regions in magnetic resonance imaging (MRI).
  • Prostate MRI segmentation is challenging due to image complexities and limitations of current segmentation techniques.
  • Existing methods often struggle with noise interference and capturing long-range contextual information.

Purpose Of The Study

  • To develop an advanced deep learning model for accurate prostate segmentation in MRI.
  • To overcome the limitations of existing methods in handling prostate MR image particularities.
  • To improve the diagnostic accuracy of early prostate cancer detection through enhanced segmentation.

Main Methods

  • Proposed a U-shaped encoder-decoder network, MM-UNet, integrating Mamba and Convolutional Neural Networks (CNN).
  • Introduced an adaptive feature fusion module with channel attention for effective hierarchical feature integration and noise suppression.
  • Developed a global context-aware module using Mamba for capturing long-range dependencies and a multi-scale anisotropic convolution module.

Main Results

  • The MM-UNet model demonstrated superior prostate segmentation performance on two public datasets.
  • Achieved state-of-the-art results, outperforming existing competing models in segmentation accuracy.
  • Validated the effectiveness of the proposed modules in enhancing feature fusion and context awareness.

Conclusions

  • The proposed MM-UNet significantly advances prostate MRI segmentation accuracy.
  • The integration of Mamba and CNN provides a powerful approach for complex medical image segmentation.
  • Future work will focus on improving model robustness and expanding its application to other medical imaging tasks.

Keywords: