成骨性肿瘤的电脑图像分析及其分类识别的实验

Computerized Image Analysis of Osteogenic Tu-mors and Their Classified Identification Test

应用电脑图像分析法，对各种不同成骨性肿瘤（良性骨瘤、骨母细胞瘤及低度、中度和高度恶性的骨肉瘤各５例）组织切片的瘤细胞密度及瘤细胞核的图像进行处理及测量，选择及提取７个特征参数（视野的细胞数，细胞核切面面积和周长的最大值、平均值及方差）。发现其中的胞核切面面积的最大值和面积的方差及核切面周长的最大值和周长的方差等随着肿瘤恶性度的增加而循序递增，呈光谱带样演变过程。而良性骨瘤与骨肉瘤之间，其瘤细胞密度，瘤细胞核面积及核周长的均数亦有不同程度的差异，故在鉴别肿瘤的良性与恶性及判别其恶性程度具有实用意义。运用树状分类器和加权最小距离判别法，对输入的８个测试样本做了分类识别的实验，结果与病理学者预定的诊断分类完全符合，说明使用图像分析法对成骨性肿瘤进行自动分类识别是可行的。

AbstractComputerized image analysis in the procession and measurement of the nuciear image of single cells and the cell number uer visual filed in tissue sections ofvarious types of osteogenic tumors have been carriedout in five cases each of osteoma,osteoblastoma,os-teosarcomas of low, moderate and high degrees of ma-lignancy respectively. Seven specific pararncters, A-max,L-max,A-men,L-men.A-vai,L-vai and cellnumber,have been chosen for comparison of the de-gree of mallgnancy in different tumors.It was foundthat A-max,L-max,A-vai and L-vai irereased gradu-ally in accordance with tumor malignancy, like a grad-uallyevoiving process of a spectrum.A-men.L-menand N-um also show differences between benign and malignant tumors. Thus, there is practical importancein differential diagniosis of tumors.Demonstrated by the binary tree classifier adn minimum range distin-guishing method to the original pathologic classifica-tion.So the classifide identification of the osteogenic tumors can be accomplished by the computerized imageanalysis.