论文部分内容阅读
目的探讨巴氏染色膀胱尿路上皮癌脱落细胞的计算机判别方法。方法取107例尿液涂片中386个正常尿路上皮脱落细胞(uroepithelium normal exfoliated cells,UNC)、439个尿路上皮结构不良脱落细胞(urothelium dysplastic exfoliated cell,UDC)和500个膀胱尿路上皮癌脱落细胞(bladder urothelial cancer exfoliated cell,UCC),将上述3种细胞利用SPSS随机分为训练组(n=1077)和识别组(n=248),测试细胞质和细胞核的色度学和几何形态参数,对训练组细胞采用逐步判别分析建立判别函数,计算回代判别准确率,然后用识别组细胞评价函数的判别效果,并计算107例标本的判别准确率。结果训练组细胞的回代判别率为80.8%,识别组细胞的判别率为80.2%,107例尿液涂片的判别率为92.5%;其判别效果明显优于单独基于色度学和几何形态参数所建立的函数(P<0.05)。结论结合色度学和几何形态参数所建立的函数对于膀胱尿路上皮癌的识别效果好。
Objective To investigate the computerized discriminating method of the exfoliated cells of the bladder pasteurea urothelial carcinoma. Methods 386 normal uroepithelium exfoliated cells (UNC), 439 urothelial dysplastic exfoliated cells (UDC) and 107 bladder urothelial cells (UCC). The three kinds of cells were randomly divided into training group (n = 1077) and identification group (n = 248) using SPSS to test the color and geometry of cytoplasm and nucleus Parameters, discriminant function was established by stepwise discriminant analysis in training group cells, and the accuracy of discriminant analysis was calculated. Then the discriminant effect of cell differentiation function was evaluated, and the discriminant accuracy of 107 samples was calculated. Results The discriminant rate of the cells in the training group was 80.8%, the discrimination rate of the cells in the identification group was 80.2%, and the discrimination rate of 107 smears in urine samples was 92.5%. The discrimination results were obviously better than those based on colorimetry and geometry alone Parameters established by function (P <0.05). Conclusion The established function combined with colorimetric and geometric parameters has a good effect on the identification of bladder urothelial carcinoma.