Enhanced Residual Network Framework for Robust Classification of Noisy Lung Cancer CT Images

Globally, lung cancer is the leading cause of cancer-related mortality. Millions of people worldwide lose their lives to cancer every year, and lung cancer is among the most common forms. Lung cancer has been successfully divided into several groups by our research. Non-small Cell Lung Cancer (NSCLC) is the more common of the two basic types of lung cancer. The ability of Convolutional Neural Networks (CNNs) to differentiate between benign and malignant tissues in CT scan pictures has been verified by recent technological breakthroughs. As part of our efforts to improve lung CT image analysis, to remove noise distortions and improve the quality of the images. This improvement is an essential precondition for accurate lung cancer identification. Add noises to the input lung CT images, such as speckle and salt-and-pepper noise, to mimic real-world situations. To determine the ideal filter, this effort will methodically assess different filtering techniques throughout a range of noise densities, from 5% to 50%. The goal of this effort is to identify lung cancer by using machine learning techniques on CT scan pictures, which will enable early and accurate cancer detection. The suggested methodology consists of two main stages: first, image classification using a Residual-based network of 152 layers of a Convolutional Neural Network, and second, noise removal using the chosen optimal filter. This network is capable of accurately classifying the benign and malignant categories found in CT scan pictures.