AN IMPROVED LUNG CANCER SEGMENTATION AND DETECTION MODEL FROM CT-SCAN IMAGES USING DEEP LEARNING

Abstract

The most promising strategy to improve a patient's chances of survival is to discover cancer early. This research uses an artificial neural network to construct a computer-aided classification approach for computed tomography (CT) images of the lungs. From the CT scan images, the complete lung is segmented, and the parameters are determined from the segmented image. Lung cancer is one of the most lethal cancers that results in a large number of deaths worldwide. The only method to increase a patient's chances of survival is to discover lung cancer early. A computed tomography (CT) scan is used to locate a tumor and determine the extent of malignancy throughout the body. In this research, we develop a model for Segmentation and Detection of Lung Cancer that is known as a SDLC System using K-means with Swarm-based Grasshopper Optimization (SGO) algorithm for CT-scan images using Deep Learning is proposed. For the autonomous lung cancer diagnosis in CT images, a novel approach is offered. The concept of Maximally Stable Extremal Regions (MSER) along with the SGO algorithm is used for feature selection. The concept of Convolutional Neural Network (CNN) is utilized to train the SDLC system on the basis of selected feature from MSER feature set via fitness function. The Optimized CNN is used to train the SDLC model using a modified K-means based Lung nodule segmentation. The accuracy of lung region segmentation is improved by employing the K-means along with the SGO algorithm. As a result, the SDLC model's classification performance has improved in comparison to previous studies in terms of precision rate and the achieved precision is near to 94.5%.