Local Ternary Co-occurrence Patterns based Lung Nodules Detection

Lung nodules are the common indicator of lung cancer. Computed tomography (CT) images are very useful method for detecting lung nodules due to its precision. The large number of slices in each CT scan needs the computer aided detection (CAD) techniques to detect the presence of nodules to assist radiologists. The aim of this paper is to detect the lung nodules using a feature extraction technique called local ternary co-occurrence patterns (LTCoP). Using LTCoP features, support vector machine (SVM) is trained and thereby detecting lung nodules and eliminating non-nodules. Comparison is made between LTCoP features and Local Binary Pattern (LBP) features in lung nodule detection. It is found that LTCoP features outperforms LBP in lung nodule detection with 91.5% accuracy.