Extraction of the Essential Region in Computer-aided Diagnosis of Gastric Atorophy Using Gastric X-ray Images

Abstract

To design a computer-aided diagnosis of gastric atrophy, this paper presents a method for extracting the essential region for the diagnosis from gastric X-ray images automatically. Although the automated extraction of gastric area in the X-ray images has a core role as a pre-processing method in the diagnosis, the extraction had been avoided in all the existing systems for diagnosing gastric abnormality because the contour of gastric area is not always clear and other objects often overlap with the contour in the X-ray images. Hence, in this paper, an extraction of only the essential region for the diagnosis is investigated. First, two objects of the barium-pool and the spinal column are extracted. Second, with the clue of these objects and difference of density distributions between the gastric area and its outer area, the essential region is extracted by detecting the contour of gastric area. Experimental results for the proposed method using 117 gastric X-ray images have shown that the proposed method correctly extracted the essential region from 95 images.