Automated classification of pneumoconiosis radiographs based on recognition of small rounded opacities

✍ Scribed by Xuan Chen; Jun-Ichiro Toriwaki; Jun-Ichi Hasegawa

• Publisher: John Wiley and Sons
• Year: 1990
• Tongue: English
• Weight: 1016 KB
• Volume: 21
• Category: Article
• ISSN: 0882-1666
• DOI: 10.1002/scj.4690211205

✦ Synopsis

Abstract

This paper proposes an automated procedure for both the size classification and the density (profusion) classification of the pneumoconiosis radiograph. The procedure is as follows. The rounded opacities are enhanced using a second‐order difference linear filter with a circular weighting function which is designed based on the model for the shape and the density distribution of the pneumoconiosis rounded opacities. The candidates for the center of the rounded opacity are extracted, and the size classification is made by estimating the size of the rounded opacity. Then the density classification is made by measuring the density of the rounded opacity. Thus, the classification procedure is conceptually composed of three stages: (1) recognition of rounded opacity; (2) size classification; and (3) density classification. This paper concentrates on the size classification and the method is described. The result is shown with a discussion for the classification experiment using ILO international standard film. In the experiment, the recognition rate of approximately 72 percent was obtained for both the size and the density classifications, using only one feature parameter. This result indicates the usefulness of the proposed method, and an attempt is made to recognize directly the individual pneumoconiosis rounded opacity. Another feature of the method is that the weighting function is selected based on the enhancement effect of the filter. This point may be interesting as an example of the automatic parameter determination of the emphasis filter in the image processing.