JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

GA-UNet: A Lightweight Ghost and Attention U-Net for Medical Image Segmentation

  • 0College of Electronics and Information Engineering, Sichuan University, Chengdu, 610065, China.
Journal of Imaging Informatics in Medicine +

|

March 13, 2024

|

March 13, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

We developed GA-UNet, a lightweight U-Net model for medical image segmentation. This efficient model achieves superior performance across multiple datasets while significantly reducing parameters and computational cost.

Area Of Science

  • Medical Image Analysis
  • Deep Learning
  • Computer Vision

Background

  • U-Net is a prominent architecture for medical image segmentation.
  • Existing U-Net variants often increase model complexity (parameters and FLOPs).
  • There is a need for efficient and high-performing segmentation models.

Purpose Of The Study

  • To introduce GA-UNet, a novel lightweight U-Net architecture.
  • To enhance feature extraction and reduce computational overhead in medical image segmentation.
  • To evaluate GA-UNet's performance against state-of-the-art models.

Main Methods

  • GA-UNet integrates lightweight GhostV2 bottlenecks to minimize redundant features.
  • Convolutional Block Attention Modules (CBAM) are employed to capture salient features.
  • The model was evaluated on four diverse medical imaging datasets: CVC-ClinicDB, 2018 Data Science Bowl, ISIC-2018, and BraTS 2018 LGG.

Main Results

  • GA-UNet achieved high F1-scores and mIoU across all tested datasets.
  • Specific results include F1/mIoU of 0.934/0.882 on CVC-ClinicDB and 0.922/0.860 on 2018 Data Science Bowl.
  • The model demonstrated superior performance compared to other state-of-the-art methods.
  • GA-UNet boasts significantly fewer parameters (2.18M) and lower FLOPs (4.45G).

Conclusions

  • GA-UNet represents a highly efficient and effective solution for medical image segmentation.
  • The proposed architecture successfully balances performance with reduced computational complexity.
  • GA-UNet offers a promising alternative for resource-constrained medical imaging applications.

Keywords: