Related Experiment Video
Updated: Jun 25, 2026
08:18
Analysis of Lymph Node Volume by Ultra-High-Frequency Ultrasound Imaging in the Braf/Pten Genetically Engineered Mouse Model of Melanoma
Published on: September 8, 2021
2.9K
SC-Unext: A Lightweight Image Segmentation Model with Cellular Mechanism for Breast Ultrasound Tumor Diagnosis
Fenglin Cai1, Jiaying Wen2, Fangzhou He1
1Department of Intelligent Technology and Engineering, Chongqing University of Science and Technology, Chongqing, 401331, People's Republic of China.
Journal of Imaging Informatics in Medicine
|February 29, 2024
Summary
This study introduces a novel segmentation model for breast ultrasound images, inspired by cellular processes. The SC-Unext model enhances accuracy while reducing computational demands, making it suitable for medical applications.
Area of Science:
- Medical Image Analysis
- Deep Learning in Radiology
- Computational Pathology
Background:
- Automatic breast ultrasound image segmentation is crucial for medical image processing.
- Existing segmentation methods face challenges with high computational complexity and large model parameters, especially for complex images.
Purpose of the Study:
- To develop a novel, efficient segmentation model for breast ultrasound images.
- To improve segmentation performance while reducing model parameters and computational cost.
Main Methods:
- The study adapted the Unext network architecture, incorporating encoder-decoder features.
- Novel apoptosis and division algorithms, inspired by cellular mechanisms, were designed and integrated.
- Spatial and channel convolution blocks were introduced into the model architecture.
Main Results:
- The SC-Unext model achieved high performance on the BUSI dataset (Dice score: 75.29%, accuracy: 97.09%) and a collected dataset (Dice score: 90.62%, accuracy: 98.37%).
- The model demonstrated efficient inference speed (92.72 ms/instance on CPU) and low resource consumption (1.46M parameters, 2.13 GFlops).
Conclusions:
- The proposed SC-Unext model significantly improves breast ultrasound tumor segmentation.
- Its lightweight design and efficiency offer substantial value for practical medical applications, particularly in resource-constrained environments.
