Prognosis Prediction in Non-Small Cell Lung Cancer using Radiomic Features Extracted from Preoperative CT Images and Transfer Learning

Abstract

Lung cancer, also known as non-small cell cancer, is a disease in which cancerous cells develop in the lung tissues and is thought to be the most dangerous to human health. It is essential to accurately detect the pathological type of this cancer before it can be treated. The most common test is a histological examination. Lung cancer, also known as non-small cell cancer, is a disease in which malignant cells develop in the lung tissues and is thought to be the most dangerous to human health. It is essential to accurately detect the pathological type of this cancer before it can be treated. A histopathological examination is the most common, habitual, and promising type of pathology examination as of today. The risk of death is reduced when the disease is detected early. In this paper, transfer learning techniques are used to study the various forms of non-small cell lung cancer. The transfer learning strategy is explored in view of the limited availability of Computed Tomography (CT) images in real life situations. An image dataset of 613 CT images is considered in this study. The networks such as AlexNet, GoogleNet and ResNet -50 are trained to identify and classify the images into normal cells, squamous cell carcinoma, adenocarcinoma, and large cell. According to the current study, AlexNet outperforms GoogleNet and Resenet-50 in the classification of non-small cell lung cancer with carcinoma, with an accuracy of 95%.

Ähnliche Arbeiten