The application of the Fuji computed radiography system for the diagnosis of pneumoconiosis, 2

This is the second part of the project shown in the title. Here we developed several image processing procedures to extract quantitative features for diagnostic aid, and to classify sizes and types of pneumoconiosis from chest X-ray films automatically. Those procedures include automated extraction of small round opacities, calculation of the density of opacities, and classification of a given film into two major categories (normal or abnormal), or into more detailed categories such as the type 0, 1, 2, and 3, and the type p, q, and r. Recognition rate was about 60 to 75%, depending on features and sample films used in the experiment. Methods to generate a sketch of a chest X-ray image by computer are also presented. (author)