Quantitative diagnosis of pneumoconiosis based on recognition ofsmall rounded opacities in chest X-ray images

摘要

A method is presented for automated categorization ofpneumoconiosis chest radiographs. The method is based on recognition ofindividual lesions in a film. The processing procedure consists of thefollowing four steps: extraction of candidate regions of small roundedopacities, recognition of abnormal opacities, calculation of the opacitydensity and film categorization. The opacity density is defined by thearea of the number of the opacities detected within a certain lungregion. In the experiments using ILO international standard films, theopacities detected by computer were highly coincident with those tracedby an experienced radiologist, and the density values calculated werealso strongly correlated with the classification code of the standardfilms