Detection of Lung Cancer Using Optimal Hybrid Segmentation and Classification

Abstract

Recently, the world health organization (WHO) showed that the lung tumor is the major leading cause of mortality. Segmentation of lung tumor is one of the exciting field for the efficient detection of lung cancer. Computerized tomography (CT) scans are used for finding the tumor position and find the cancer level in the body. This work introduces an automated diagnosis model for increasing the survival rate of patients. This work undergoes the major stages like pre-processing, hybrid segmentation, feature extraction and classification. The artifacts in the input CT images are pre-processed using noise removal technique. Then, the pre-processed image is subjected to the segmentation process. Here, the segmentation is carried out by Hybrid optimal clustering and improved region growing algorithm (IRGA). Hybrid optimal clustering is the integration of fuzzy C means clustering (FCM) and the optimization Harris Hawk algorithm (HHA). Finally, the deep features from the CT lung images are extracted and classifiedby the deep learning (DL) model squeezenet. The proposed model is tested on the benchmark dataset Early Lung Cancer Action Program (ELCAP) and achieved better accuracy and specificity of 0.996 and 0.992 respectively.