Mattsson U, Chodorowski A, Gustavsson T, Heyden G, Jontell M, Bergquist F
Department of Oral Pathology, Faculty of Odontology, Göteborg University.
Oral Surg Oral Med Oral Pathol Oral Radiol Endod. 1995 Feb;79(2):199-206. doi: 10.1016/s1079-2104(05)80282-3.
The objective of the present study was to elaborate and to evaluate a method in which the clinical appearance of a mucosal lesion could be expressed by means of color and morphologic features extracted from digitized images. The ability to express mucosal changes with image features was evaluated, as was the accuracy of a classification of the lesions on the basis of image feature analysis. Computer-assisted image analysis was performed on digitized color slides from 76 patients with lichenoid reactions that included both oral lichen planus and contact lesions and 20 patients with homogenous oral leukoplakia. Color features were measured according to the intensity-hue-saturation color system, and morphologic features were estimated by tracing the hyperkeratotic area. Hyperkeratinizing and inflammatory tissue reactions could be demonstrated and defined by image analysis. Image features indicating inflammatory tissue reactions were more pronounced in cases of oral lichenoid reactions. Morphologic parameters could express the anatomic appearance of the lesions. The comparison between histopathologic diagnosis and a classification through computerized analysis demonstrated that 90.5% of oral lichenoid reactions could be separated from homogenous oral leukoplakia by linear discriminant analysis.
本研究的目的是阐述并评估一种方法，通过该方法可利用从数字化图像中提取的颜色和形态特征来表达黏膜病变的临床表现。评估了用图像特征表达黏膜变化的能力，以及基于图像特征分析对病变进行分类的准确性。对76例苔藓样反应患者（包括口腔扁平苔藓和接触性病变）以及20例均质型口腔白斑患者的数字化彩色玻片进行了计算机辅助图像分析。根据强度-色调-饱和度颜色系统测量颜色特征，并通过描绘角化过度区域来估计形态特征。通过图像分析可以显示和定义角化过度和炎症组织反应。在口腔苔藓样反应病例中，表明炎症组织反应的图像特征更为明显。形态学参数可以表达病变的解剖学外观。组织病理学诊断与通过计算机分析进行的分类之间的比较表明，通过线性判别分析，90.5%的口腔苔藓样反应可以与均质型口腔白斑区分开来。
