Related Experiment Video
Updated: Mar 28, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
859
SECT-Net: hybrid dual-decoder network with SE-convolution transformer for liver tumor segmentation
Qiquan Zeng1, Dongfen Ye2, Meiqin Chen3
1College of Mechanical Engineering, Quzhou University, Quzhou, China.
Frontiers in Physiology
|March 27, 2026
Summary
This study introduces SECT-Net, a novel hybrid dual-decoder network for accurate liver tumor segmentation. SECT-Net effectively addresses segmentation challenges, demonstrating robust performance on diverse datasets.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Accurate liver tumor segmentation is crucial for patient prognosis and treatment planning.
- Automated segmentation faces challenges due to tumor heterogeneity, indistinct boundaries, and varied appearances.
Purpose of the Study:
- To develop an advanced deep learning model for precise liver tumor segmentation.
- To overcome limitations of existing methods in segmenting complex liver tumors.
Main Methods:
- Proposed SECT-Net, a hybrid dual-decoder network integrating squeeze-and-excitation convolution (SE-convolution) and Transformer attention.
- Incorporated SE-convolution Transformer modules (SECTM) and a deep feature capture module (DFCM) into an encoder-decoder architecture.
Main Results:
- SECT-Net achieved high segmentation performance on private datasets (Dice: 0.8452 arterial, 0.8425 portal venous) and a public dataset (Dice: 0.8845).
- Demonstrated robust generalization capabilities with consistent Dice, Mcc, and Jaccard scores across different datasets.
Conclusions:
- SECT-Net shows strong reliability and robustness for liver tumor segmentation.
- The proposed network effectively handles diverse intensity distributions and morphological characteristics of liver tumors.
