Detection of Pneumoconiosis Opacities by Local Pattern Processing and Its Application to Automatic Diagnosis
Toshihiro Watanabe, Hidefumi Kobatake · Transactions of the Society of Instrument and Control Engineers · 1989
This paper presents a study on the image processing method for recognizing small rounded opacities of pneumoconiosis in a digitized chest X-ray image and an automatic diagnostic system is also described.A local processing filter which detects candidates of small rounded opacities of pneumoconiosis is proposed. This filter evaluates local density pattern of lung area and discriminates small rounded opacities from otheri rregular ones. After applying the local processing filter, revaluation of the density pattern of the neighboring region of each candidate is performed. Candidates whose neighboring density patterns are not consistent with those of small rounded opacities of pneumoconiosis are purged and the final results are obtained.The latter half of this paper concerns with the automatic diagnosis of pneumoconiosis. The number of detected opacities per unit lung area, the total sum of their areas and so on are used as features for the categorization of profusion of opacities. Experiments to test the performance of the proposed system have been performed.Zonal decision rates for normal/abnormal classification and for the four major category classification were 82 and 69 percent, respectively. Rates for overall film decision based on the majority rule were much higher than those rates. These experimental results show the effectiveness of the proposed system.