An Effective Method for Lung Tumor Screening Using CT Dataset

Abstract

Lung tumors are one of the most dangerous forms of cancer. It has a high incidence and mortality rate because it is frequently found at a later stage. Computed tomography (CT) scans are frequently used to distinguish between illnesses. Computerized systems have been created to analyze disease in its early phases. This paper describes a completely automated framework for detecting nodules in lung CT images. Grayscale CT image histograms are computed to automatically separate lung regions from the underlying tissue. Morphological operators are used to refine the output. The internal anatomy of the parenchyma should then be extracted. In order to differentiate candidate nodules from other structures, a threshold-based technique has been suggested. For these node candidates, various statistical and shape-based features are extracted to create a node feature vector that is classified using a support vector machine. The proposed method is tested on a large lung CT data collection gathered by the Lung Imaging Database Consortium. (LIDC). When compared to comparable existing methods, the proposed strategy produced better results. Its efficacy has been demonstrated by a sensitivity rate of 84.6%.