Multiple Types of Cancer Classification Using Histopathological Images Based on Learning Without Forgetting Powered Deep Learning Models

Cancer is the second biggest cause of death worldwide, accounting for one of every six deaths. However, the odds of survival are greatly increased by early disease detection. We may be able to assess more cases in less time if we automate cancer diagnosis using artificial intelligence (AI). This study suggests using AI-based deep learning models to categorize photos of eight different cancer types, including breast, cervical, brain, and lung cancer. Convolutional Neural Networks (CNN), one type of deep learning model, are assessed in this paper for their ability to classify photos with characteristics of cancer. To recognize various types of cancer cells, pre-trained CNN variants like MobileNet, VGGNet, and DenseNet are used to apply the information they acquired using the ImageNet dataset. To determine the appropriate values for the hyperparameters, we employ Bayesian Optimization. Transfer learning, however, has the potential to render models incapable of classifying the original datasets on which they were trained. Thus, we employ Learning without Forgetting (LwF), which preserves the network’s inherent capabilities while training the network solely with fresh task data. The experiments’ findings demonstrate that the suggested transfer learning-based models outperform the state-of-the-art methods in terms of accuracy.