Advancements in Breast Cancer Detection: A Holistic Evaluation of Deep Learning Models with Histology and Thermal Imaging Datasets

This research addresses the pressing issue of breast cancer detection, emphasizing the development and evaluation of deep learning models using two distinct datasets. The first dataset involves histology images, where Convolutional Neural Networks (CNN), ResNet50, and a novel Concatenated Model are employed and rigorously evaluated. The second dataset comprises thermal images, providing an alternative perspective for breast cancer detection. The study reveals promising outcomes, with the Concatenated Model demonstrating the highest overall performance on histology images, achieving an accuracy of 87%, precision of 77%, recall of 77%, and an F1-Score of 77%. For thermal images, CNN achieves an accuracy of 80%, albeit with a lower recall of 51%, while ResNet50 attains the highest recall at 67%. These findings underscore the efficacy of deep learning models for breast cancer detection and emphasize the importance of considering diverse datasets and modalities for a comprehensive evaluation, contributing valuable insights to developing robust breast cancer screening systems to enhance early detection rates and save lives.