A Computer-Aided Early Detection System of Pulmonary Nodules in CT Scan Images

Abstract

In the present paper, computer-aided system for the early detection of pulmonary nodules in Computed Tomography (CT) scan images is developed where pulmonary nodules are one of the critical notifications to identify lung cancer. The proposed system consists of four main stages. First, the raw CT chest images were preprocessed to enhance the image contrast and eliminate noise. Second, an automatic segmentation stage for human's lung and pulmonary nodule candidates (nodules, blood vessels) using a two-level thresholding technique and a number of morphological operations. Third, the main significant features of the pulmonary nodule candidates are extracted using a feature fusion technique that fuses four feature extraction techniques: the statistical features of first and second order, Value Histogram (VH) features, Histogram of Oriented Gradients (HOG) features, and texture features of Gray Level Co-Occurrence Matrix (GLCM) based on wavelet coefficients. To obtain the highest classification accuracy, three classifiers were used and their performance was compared. These are; Multi-layer Feed-forward Neural Network (MF_NN), Radial Basis Function Neural Network (RB-NN) and Support Vector Machine (SVM). To assess the performance of the proposed system, three quantitative parameters were used to compare the classifier performance: the classification accuracy rate (CAR), the sensitivity (S) and the Specificity (SP). The developed system is tested using forty standard Computed Tomography (CT) images containing 320 regions of interest (ROI) obtained from an early lung cancer action project (ELCAP) association. The images consists of 40 CT scans. The results show that the fused features vector which resulted from GA as a feature selection technique and the SVM classifier gives the highest CAR, S, and SP values of99.6%, 100% and 99.2%, respectively.