论文部分内容阅读
应用自动图像分析的方法，对内镜下检出的大肠隆起性病变１８３例活检标本的细胞核形态特征进行定量研究。结果显示非腺瘤性息肉、腺瘤及腺癌细胞核形态的各项参数值均随病变恶性程度的增加而增大，参数值在正常及各项隆起性病变组间的差异具显著性（Ｐ＜０．０１）。将上述数据输入计算机进行多因素逐类判别并建立函数判别式，通过回代矩阵分析与原诊断的符合率为８９％，其中大肠癌、腺瘤、非腺瘤性息肉分别为８６％、８５％、９５％，显示出一定的鉴别诊断功能和临床应用价值。
Using automated image analysis methods, the morphological features of 183 cases of biopsy specimens detected by endoscopic examination of the large intestine uplift lesions were quantified. The results showed that the values of nuclear parameters of non-adenomatous polyps, adenomas and adenocarcinoma cells increased with the increase of the degree of malignancy, and the differences between the parameters in the normal and various prominence groups were significant (P <0.01). The above data was entered into a computer for multi-factorial class discrimination and a function discriminant was established. The coincidence rate with the original diagnosis was 89%. The colorectal cancers, adenomas, and non-adenomatous polyps were 86% and 85, respectively. %, 95%, showing a certain differential diagnosis function and clinical application value.