Usage of Template Matching for Lung Nodule Detection in CT Images

In this paper we describe the lung nodule detection through image processing methods. Therefore, in the preprocessing stage, we use a medical-like method to omit all objects that surely they are not nodule. In this way nodule candidates are obtained. Therefore, false positives decay, while rate of main process does not increase. Finally, nodule candidates are selected through these suspicious regions. In the first step, we extracted candidate areas of the lung using synchronous thresholding in consecutive slices, medical decision implementation and also morphology methods. We divided the preprocessing stage into two phases for detecting all of lung nodules more accurately, depend on the nodule is connected to lung wall or vessel or it is an alone nodule.Here, template matching and maximum similarity method have used for nodule detection. The used database comes from LIDC database images that consist of 7 patients' CT scans. All of nodules are completely detected and there are 3 FP/slice. Value of similarity has been computed for all of points (pixels) that determined in preprocessing stage. Indeed, the limited search area makes the optimum (i.e. direct search) algorithm as fast as other sub-optimal search methods. Indeed, this method is faster than other methods that don’t have enough attention to information of CT scan slices.