Description of Invention
Breast cancer, a leading cause of death among women, requires precise evaluation of HER2 receptor status to determine prognosis and better treatment. This project introduces an automated HER2 IHC scoring method using deep learning, specifically a novel approach based on the High-Resolution Network (HRNet) architecture. The method utilizes the public dataset of 3,656 images and a private dataset of 75 whole slide images for real-time performance assessment. The model was able to achieve 90% accuracy which enhancing diagnostic reliability. The system's ability to provide confidence scores and reduce subjectivity suggests its potential to improve breast cancer diagnosis and patient care.