MULTIPLE CANCER TYPES CLASSIFIED USING CTMRI IMAGES BASED ON LEARNING WITHOUT FORGETTING POWERED DEEP LEARNING MODELS

Abstract

We recommend using AI, principally deep learning models, to consequently analyze lung, brain, breast, and cervical cancer. We use CNNs like VGG16, VGG19, DenseNet201, MobileNetV3 (both small and big variations), Xception, and InceptionV3 with transfer learning from pre-prepared models like MobileNet, VGGNet, and DenseNet. Bayesian Advancement enhances hyperparameters for model execution. Learning without Forgetting (LwF) holds network abilities while further developing order exactness on new datasets to conquer transfer learning hardships. We found that MobileNet-V3 little has 86% accuracy on the Multi Cancer dataset, beating different strategies. Expectation procedures utilizing Xception and InceptionV3 are investigated to further develop execution to 90% or higher. We likewise propose a Flask module to build an easy to use front-end for verification based client testing. This study shows that AI-driven cancer detection could improve early determination and treatment.

INDEX TERMS Cancer, convolutional neural network (CNN), pretrained models, Bayesian optimization,transfer learning, learning without forgetting, VGG16, VGG19, DenseNet, mobile net