Related Experiment Video
Updated: Sep 17, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
529
DynTransNet: Dynamic Transformer Network with multi-scale attention for liver cancer segmentation
Siming Zheng1,2, A S M Sharifuzzaman Sagar3, Yu Chen4
1The First Affiliated Hospital of Fujian Medical University, Fuzhou, China.
Frontiers in Oncology
|July 4, 2025
Summary
This study introduces an advanced U-shaped framework for accurate liver tumor segmentation in medical scans. The novel approach improves diagnostic accuracy for hepatocellular carcinoma (HCC), aiding clinical decision-making.
Area of Science:
- Medical Image Analysis
- Artificial Intelligence in Healthcare
- Oncology
Background:
- Hepatocellular carcinoma (HCC) is a leading cause of cancer mortality.
- Accurate liver tumor segmentation in CT/MRI is crucial for treatment planning.
- Manual segmentation is time-consuming, error-prone, and inconsistent.
Purpose of the Study:
- To develop a novel, robust, and accurate automated framework for liver tumor segmentation.
- To address challenges in boundary precision, complex structures, and dataset imbalance.
Main Methods:
- A U-shaped segmentation framework inspired by U-Net.
- Dynamic Multi-Head Self-Attention (D-MSA) in the encoder for spatial dependencies.
- Skip connections in the decoder, a Feature Mix Module (FM-M), and a Residual Module (RM).
Main Results:
- Achieved superior segmentation performance on benchmark datasets.
- Mean Dice scores of 86.12% on ATLAS and 93.12% on LiTS datasets.
- Demonstrated enhanced accuracy and robustness in liver tumor delineation.
Conclusions:
- The proposed framework offers a robust and efficient tool for automated liver tumor segmentation.
- Potential to streamline diagnostic workflows and improve medical image analysis.
- Significant advancement in computer-aided diagnosis for HCC.
