Detection and segmentation of nodules in chest radiographs based on lifetime approach

Abstract

Early detection and treatment opportunities for lung cancer is reduced the mortality of this disease. Chest radiography is one of the commonly used screening methods for the preliminary diagnosis of lung cancer. In this study, an algorithm for nodule detection in chest radiograph image is presented. It takes into account the suspicious salient regions. Firstly, in order to enhance the image contrast, the CLAHE filter is applied. Then, local maximal regions are extracted by multi-scale approach based on optimum lifetime. Some of these regions are eliminated by decision tree using the morphologic and the intensity features for detection and segmentation of candidate nodules. Finally, the texture features extracted from the segmented regions are classified by using RusBoost method. The method has been tested on the JSRT (Japanese Society of Radiological Technology) database images. Experimental results demonstrate that the proposed method achieves a very satisfactory performance for detection and segmentation of the suspicious salient regions at the same time.