Hybrid CNN and RNN Model for Histopathological Sub-Image Classification in Breast Cancer Analysis Using Self-Learning

Breast cancer, a pervasive and life-threatening disease, necessitates the development of advanced classification techniques. This paper introduces a model that combines Convolutional Neural Networks with Recurrent Neural Networks to classify sub-images in breast cancer. By leveraging localized features from a pre-trained CNN and insights from the RNN, this innovative approach aims to enhance accuracy. A sub-image-based strategy is employed to capture localized characteristics more effectively. A hierarchical self-learning approach is implemented to gradually correct mislabeled images, utilizing an invariant rule informed by prior knowledge of potential labeling errors. The model incorporates VGG19, Google Net, and ResNet101 for classifying breast cancer sub-images at various magnifications (40X, 100X, 200X, and 400X) from the BreaKHis dataset. Among these, ResNet101 demonstrates a notable classification accuracy of 98.58% with CNN techniques. However, the hybrid model achieves an impressive accuracy of 99.76%. This approach is promising for advancing medical image classification, offering potential diagnosis and patient care improvements.