Automatic diagnosis of pneumoconiosis by texture analysis of chest X-ray images

This paper presents a study on automatic diagnosis of pneumoconiosis by texture analysis of chest X-ray images. A pre-processing method for normalizing distribution of film density is proposed and its effectiveness is studied experimentally. Dominant causes of the deviations of texture features for evaluating the profusion of small opacities are excluded by the pre-processing. New texture features based on the distribution of gradient vectors are also proposed. Experiments to evaluate the proposed system are demonstrated.