Advancing Breast Cancer Prediction and Early Detection with Advanced Deep Learning Models

Breast cancer is a common disease that predominantly affects women worldwide, with the potential to be fatal. Histopathologists utilize various criteria to examine tissue samples under a microscope for diagnosing cancer through histopathological imaging. However, pathologists often disagree during the diagnostic process. To address this issue, Deep Neural Networks are employed for supervised classification, given the considerable time required for manual categorization of histological pictures. In our classification task, we utilized the Breast Histology dataset, comprising 241 training and 21 test pictures. Effective preprocessing of the dataset is crucial for successful categorization. Images were classified into four categories (Normal, Benign, Insitu carcinoma, and Invasive cancer) using transfer learning based on GoogleNet, AlexNet, and ResNet. Notably, our method achieved the highest accuracy, with ResNet reaching 97.11%. Ongoing research aims to enhance the technique's effectiveness and reduce reliance on human involvement. Furthermore, this promising approach can be adapted to automate additional medical imaging modalities, potentially advancing the automation of various aspects of medical diagnosis through improvements to the suggested framework.