Deep Learning Analysis of Histopathology Images for Breast Cancer Detection: A Comparative Study of ResNet and VGG Architectures
Andreas Kanavos, Phivos Mylonas
Abstract
Medical image analysis has undergone significant advancements with the emergence of deep learning techniques, offering great promise in improving diagnostic precision and expediting patient care. This research investigates the effectiveness of ResNet and VGG architectures in detecting breast cancer through the analysis of histopathology images. By meticulously fine-tuning hyperparameters and optimizers, we establish robust and accurate deep learning models. Our findings reveal that the ResNet model with the SGD optimizer excels, surpassing the performance of VGG in terms of accuracy and F1-score. However, employing transfer learning with pre-trained VGG16 and ResNet50 networks does not yield competitive results, potentially due to disparities in input image size and data distribution. The primary focus of this study is to address the critical challenge of early breast cancer detection, ultimately leading to enhanced patient outcomes. By exploring state-of-the-art deep learning architectures and methodologies, we contribute to the growing body of research aimed at leveraging artificial intelligence for medical diagnosis.