Related Experiment Video
Updated: Sep 11, 2025
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
C5-net: Cross-organ cross-modality cswin-transformer coupled convolutional network for dual task transfer learning in
Meng Wang1, Haobo Chen1, Lijuan Mao2
1The SMART (Smart Medicine and AI-based Radiology Technology) Lab, Shanghai Institute for Advanced Communication and Data Science, Shanghai University, Shanghai, China; Key Laboratory of Specialty Fiber Optics and Optical Access Networks, School of Communication and Information Engineering, Shanghai University, Shanghai, China.
We developed C5-Net, a novel deep learning model for enhanced lymph node ultrasound diagnosis. It improves segmentation and classification accuracy, aiding in early detection of malignancy.
Area of Science:
- Medical imaging
- Deep learning
- Oncology
Background:
- Deep learning for lymph node ultrasound diagnosis faces challenges with limited data, learning local/global features, and collaborative segmentation/classification.
- Existing methods struggle with comprehensive analysis and synergistic learning for accurate diagnosis.
Purpose of the Study:
- To propose C5-Net, a novel deep learning network for improved lymph node ultrasound image segmentation and classification.
- To address data scarcity and enhance feature learning by leveraging cross-organ, cross-modality transfer learning.
Main Methods:
- Developed the Cross-organ Cross-modality Cswin-transformer Coupled Convolutional Network (C5-Net).
- Employed a transfer learning strategy using skin lesion dermoscopic images for lymph node ultrasound images.
- Coupled Transformer and convolutional neural networks for integrated local and global feature learning.
- Utilized shared encoder weights for synergistic segmentation and classification tasks.
Main Results:
- C5-Net achieved a Dice coefficient of 0.854 for segmentation and 0.874 for classification accuracy.
- The model demonstrated superior performance compared to advanced methods in lymph node segmentation and classification.
- The study utilized 690 lymph node ultrasound images and 1000 skin lesion dermoscopic images.
Conclusions:
- C5-Net effectively addresses key challenges in lymph node ultrasound diagnosis.
- The proposed method shows high accuracy and robustness, facilitating early and accurate detection of lymph nodal malignancy.
- This contributes to improved treatment planning in clinical oncology.
