Objective Quantification of Grade of Atypia in Epithelial Tumors of the Stomach by Image Processing

Abstract

An image processing technique to extract tubular and nuclear areas from histological specimens stained with hematoxylin was established. Two indices representing tubular density and nucleus‐to‐cytoplasm ratio of glands were examined in intestinal‐type tumor of the stomach. In specimens clearly stained with hematoxylin, significant differences of brightness among nuclei, cytoplasm and stroma can be obtained. Image processing by computer enabled us to identify nuclei, cytoplasm and stroma by utilizing the differences of brightness among them. Both indices tended to increase with severity of histological atypia and their values in the tubular adenoma group were significantly different from those in the tubular adenocarcinoma group (P< 0.05). Therefore, these two indices could be regarded as valid and objective measures of the grade of histological atypia. Eighty‐four percent of tubular adenoma and tubular adenocarcinoma cases were classified correctly with the discriminant formula and critical value calculated from the two indices on the basis of measurement of the images magnified by 40 times. There is, however, still a non‐negligible overlap of discriminant scores between the tubular adenoma group and tubular adenocarcinoma group. More indices representing histological atypia will be required to allow objective differential diagnosis between malignant and benign lesions.