Detection and Classification of Lung Tumors using Machine Learning Algorithms

Several medical fields have been using image processing techniques in recent years to improve diagnosis and treatment in the early stages. It is particularly important to diagnose a patient's illness as soon as possible when the patient has multiple tumors, such as breast and lung cancers. As part of this procedure, the lungs are divided into segments and inspected individually for nodules in order to determine the disease's extent. Since CT lung tumors are known to have a variety of shapes and hazy tumor boundaries, stimulated tumor segmentation is always appealing, even though various lung tumor segments have been documented. Using high-resolution microtomography (micro-CT), it has been demonstrated that lung tumors can be detected both early and late in the formation process in living mice. The prevalence of adenocarcinoma and squamous cell lung cancer in all lung tumors is thought to be 30%• In this innovative concept, the usefulness of 2D and 3D imaging features for tumor classification is examined. For tumor classification models, wrapper and Relief-F feature selection strategies were combined with decision trees and support vector machines. The accuracy of 2D elements is almost as good as 3D elements, although 3D elements are easy to use. Visual inspection of postmortem inflated and removed lungs further validated the micro-CT results. Micro-CT technology offers a wide range of possibilities for non-invasive imaging of tumor development and progression. Micro-CT technology allows continuous lung cancer treatment research to take place.