CANCER DETECTION AND CLASSIFICATION THROUGH TRANSFER LEARNING WITH RESNET18-BASED CONVOLUTIONAL NEURAL NETWORK

Abstract

Cancer is a disease where some cells in the body grow destructively and, in most cases, spread to other body organs. Biologically, cell grows and expand through some natural processes to create new cells that replaces old and damaged ones, this process when distorted can result in abnormal cell growth thereby forming tumors that can harmful. Recently, the rapid advancements in deep learning have proffer solutions to the deadly disease by way of early detection and accurate classifications. This study explores the application of deep learning frameworks, specifically PyTorch, in developing a robust multi-cancer detection and classification system. The research focuses on creating an architecture capable of accurately diagnosing multiple cancer types, including brain, breast, bone, kidney, oral, lung/colon, cervical, and lymphoma cancers. The study employs a comprehensive methodology of obtaining large size of different cancer datasets which was preprocessed to ensure a clean, noiseless datasets, models development, and rigorous training and testing of the datasets and finally performance evaluation of the system were carried out. The results demonstrated significant improvements in diagnostic accuracy, highlighting the potential of machine learning in transforming cancer care. The proposed system offers promising prospects for enhancing early detection and personalized treatment, ultimately improving patient outcomes.