Investigation of Segmentation Classification and Severity Identification of Lung Cancer with U Shaped Network Transformer Architecture

Abstract

Lung cancer is the deadliest disease in the world. Lung diseases are a newlinesignificant concern for individuals globally. TB, Cancer, COVID-19, and Pneumonia can newlinebe diagnosed by segmenting, analyzing, classifying, and evaluating lung images of chest newlineCT scans. The manual segmentation of medical images requires much work from newlineradiotherapists. The proposed U-shaped Network Transformer (UNT) method provides a newlinepowerful mechanism for the prior prognosis as well as treatment of lung tumors, along newlinewith Tuberculosis, Pneumonia, and COVID-19. The model was trained using 2D CT newlineimages, and the performance was tested with and without the optimizers, namely ResNet- newline50, Inception-ResNet as well as VGG-19. The experimental results yielded efficient newlineperformance with 96.9% accuracy. newlineFurther, to improve prediction accuracy, an intelligent healthcare system newline(IHS) was proposed to analyze and prognose the severity of lung disease. The feature newlineextraction has been done via TL with several pre-trained DL architectures (Inception newlineVGG-19, ResNet-50, ResNet-v2) and an Ensemble Transfer Learning (ETL) Model. The newlineproposed approach has diagnosed five classes, each with 0.982 recall, 0.978 precision, newline0.974 F1-score, and 0.986 accuracy newline