Efficient Convolution Network to Assist Breast Cancer Diagnosis and Target Therapy
Ching-Wei Wang1, Kai-Lin Chu1, Hikam Muzakky1
1Graduate Institute of Biomedical Engineering, National Taiwan University of Science and Technology, Taipei 106335, Taiwan.
Cancers
|August 12, 2023
Summary
A novel deep learning method enhances breast cancer diagnosis and HER2 gene amplification detection. This AI approach offers high accuracy and efficiency, making it suitable for clinical use.
Area of Science:
- Oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Breast cancer is a leading cause of death in women globally.
- Accurate detection of human epidermal growth factor receptor-2 (HER2) gene amplification is crucial for targeted therapy.
- Current manual microscopic analysis for HER2 detection is subjective and lacks reproducibility.
Purpose of the Study:
- To introduce an efficient deep learning framework for breast cancer diagnosis and HER2 amplification detection.
- To address the limitations of manual slide analysis in pathology.
Main Methods:
- Development of a low-computing-cost deep learning method.
- Application of the framework to breast cancer diagnosis and HER2 amplification detection on Fluorescence in situ hybridization (FISH) and Dual in situ hybridization (DISH) slides.
Main Results:
- The proposed deep learning framework achieved high precision and recall in clinical applications.
- The method demonstrated superior performance compared to benchmark methods in Intersection over Union (IoU).
- Significant reductions in AI training time (16.93%), AI inference time (17.25%), and memory usage (18.52%) were observed.
Conclusions:
- The deep learning method provides accurate and reproducible results for breast cancer diagnosis and HER2 amplification detection.
- The framework's efficiency and reduced resource requirements make it clinically feasible.
