Transfer Learning Approach for Detecting Breast Cancer Tumour Subtypes in Whole-Slide Images

Breast cancer remains a significant global health concern, and accurate sub-typing of breast cancer tumours is essential for personalized treatment strategies. Whole-Slide Imaging (WSI) technology has emerged as a promising approach for comprehensive tumour analysis. In this research paper, we propose a transfer learning approach for detecting breast cancer tumour subtypes in WSI and Region of Interest (ROI). The dataset used in this study is taken from BRACS which has 547 labelled WSI images and 4539 ROI images for 189 patients. We leverage the power of pre-trained deep learning models, including InceptionV3, ResNet50, MobileNet, and VGG16, to extract high-level features from WSI images. Custom top layers are incorporated to fine-tune the models for the specific task of tumour subtype classification. InceptionV3 emerges as the most compelling model for WSI and ROI with an average F1-score of 66% and 76% respectively. Extensive experiments on a diverse dataset of breast cancer WSI demonstrate the efficacy of the proposed approach, achieving superior performance compared to traditional methods. Our findings showcase the potential of transfer learning to improve the accuracy and efficiency of tumour subtype detection in WSI and ROI, paving the way for advancements in breast cancer diagnosis and personalized treatment planning. This research contributes to the growing field of medical image analysis and highlights the significance of leveraging transfer learning for accurate and scalable tumour subtyping in breast cancer research.