Lung Nodule Detection Using Fuzzy-Tsallis Entropy and SVM

Lung nodules are the primary tissue of lung cancer. Designing a computer-aided diagnosis system can be used in order to enhance the accuracy of the radiologist in the detection and description of these tissues. The aim of this study is to provide an approach to detection of lung nodules on CT scan images. This study was conducted in two main stages of nodule segmentation and detection.  In the segmentation stage, combination procedures such as Fuzzy-Tsallis entropy and thresholding were used. In the nodule detection stage, intensity and geometric features are extracted and suspicious areas are determined by support vector machine. Using intensity features causes increasing true positive rate while using geometrical features causes decreasing false positive rate. To evaluate the proposed method, images of LIDC and Taba datasets were used. Classification sensitivity is 92%. The obtained results comparing to those reported by other articles, indicate the usefulness of the research.