Lesion detection in lung regions that are segmented using spatial relations

Abstract

In this paper, we have described a lesion detection approach from chest radiography. We have illustrated the importance of accurate segmentation as a preprocessing step in a CAD scheme. Then, a suitable combination among 118 features has been identified using the forward stepwise selection method. The main idea is to obtain a set of features that is enable a CAD not to discriminate between normal lesions and abnormal ones but to specify its nature if this lesion is an infection for example.