小儿急性白血病骨髓有核细胞的形态定量分析

The Morphologic Quantitative Analysis of Nucleated Cells on Bone Marrow Smear of Childhood AcuteLeukemis

应用计算机图像分析系统对３９例小儿急性白血病及８例对照组骨髓涂片中的有核细胞进行形态定量测定。选择１１个 参数作各不同组间的差别显著性检验，采用Ｂａｙｅｓ准则逐步判别分析，建立判别函数式。结果：（１）除细胞形状参数外，其余１０个参数（涉及细胞面积、核面积、核面积与细胞面积比、胞浆面积及核的形状）在对照组与白血病之间及各白血病亚型之间均存在着不同程度的显著性差别．（２）３组判别函数式（对照组、ＡＬＬ、ＡＭＬ间的判别；Ｌ１、Ｌ２、Ｌ３间的判别；Ｍ２、Ｍ３、Ｍ３０、Ｍ５６、Ｍ６间的判别）具有满意的判别分类效果，四代准确率均在９０％以上。

morphologic quantitative analysis of nucleated cells on bone marrow smear of 39 children withacute leukemia and 8 normal subjects as controls was Performed by using computer--assisted imageanalysis system. Eleven perameters were used for the test of significance of difference, and further,the discriminant functions were set up by Bayes stepwise discriminant analysis. The Purpose of thisPreliminary work was to explore the automated diagnosis and classification method of acute leukemiawith satisfied objectivity and repeatability. The results are as follows:1. Except for the shape parameter of cells ten parameters(relating to cell area, nuclear area, theratio between nuclear area and cell area, cytoplasmic area and nuclear shape) shoWed the significance of difference to varying degress between control and leukemia groups, and among anbtypes ofleukemias.2. The three groups of discriminant function(including discriminations among control group,ALLand AML; among L1,L2 and L3 of ALL;among M2a,M3a,M3b,M56 and M6 of AML) seemed to havesatisfied classifcation results. The rates of correctly calssified cases by means of back substitution wereall above 90%. It is indicated that automated comPuter diagnosis and classification of acute leukemiaare feasible.