Related Experiment Video
Updated: Jul 15, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
A cell image segmentation method based on edge feature residual fusion.
Jinlian Du1, Yanqiu Zhang1, Xueyun Jin1
1College of Computer Science, Faculty of Information Technology, Beijing University of Technology, Beijing, China.
Methods (San Diego, Calif.)
|September 29, 2023
Summary
This study introduces ERF-TransUNet, a deep learning model for accurate cell image segmentation. It improves cancer diagnosis by enhancing cell contour positioning and segmentation accuracy.
Area of Science:
- Medical Imaging
- Computational Pathology
- Artificial Intelligence in Medicine
Background:
- Cancer diagnosis and grading rely heavily on cellular morphology.
- Deep learning-based automatic cell segmentation is crucial for computer-aided pathological diagnosis.
- Existing models struggle with rough boundaries and inaccurate segmentation in cell images.
Purpose of the Study:
- To design an improved cell image segmentation network, ERF-TransUNet.
- To address limitations in boundary accuracy and segmentation precision.
- To leverage edge and object features for enhanced cell contour positioning.
Main Methods:
- A hybrid Convolutional Neural Network (CNN) and Transformer architecture extracts multi-scale features.
- Independent edge feature extraction modules are incorporated.
- Residual fusion modules enhance edge feature extraction and their integration with object features.
Main Results:
- ERF-TransUNet demonstrated superior performance on CRAG and Glas gland cell datasets.
- The model achieved significant improvements in Dice coefficient and Hausdorff distance compared to existing methods.
- Enhanced extraction and fusion of edge features led to more accurate cell contour localization.
Conclusions:
- ERF-TransUNet effectively improves cell image segmentation accuracy.
- The proposed model offers a promising advancement for computer-aided pathological diagnosis.
- Accurate cell segmentation is vital for reliable cancer diagnosis and grading.
