Breast Cancer Histopathological Image Classification Using PIP-Net

Histopathological images serve as pivotal assets within the domain of breast cancer diagnosis, demanding profound comprehension for precise interpretation. This paper introduces a histopathological image classification framework employing the Patch-Based Intuitive Prototype Network (PIP-Net). PIP-Net finds important features without the need for manual part annotations by using prototype patches and self-supervised learning mechanisms. Consequently, PIP-Net adeptly discerns novel patterns within images, thereby facilitating informed therapeutic interventions. Leveraging the ‘BrakeHis’ dataset, encompassing histopathological images of breast cancer, this paper evaluates PIP-Net's efficacy across varying magnification levels, specifically at 40x, 100x, 200x, and 400x. The classification accuracies achieved are 98.49%, 94.53%, 99.33%, and 98.08% for each respective magnification level. The integration of prototypical elements within images serves to enhance correlation with human visual perception. This shows the paper's dual commitment to not only attaining superior classification performance but also aligning with interpretability and intuitive principles integral to medical image analysis.