A computer aided diagnosis method for SARS in its early stage based on chest image of digital radiograph

摘要

The aim of this paper is to develop a computerized scheme in order to help clinicians to detect severe acute respiratory syndrome lesions in its early stage based on digital chest radiographs. After decomposition of the original chest radiographs with the theory of locally orderless images, the suspicious SARS lesions can be picked out. Together with some other images processing techniques to improve the accuracy, the sensitivity of this system achieved 77.8% at a coarse synthetical specificity of 22%.