乳腺癌细胞核形态参数计算机识别系统的开发研究

Correlation of Nuclear Morphometry with Progression of Breast Cancer Using Computer Image Analysis

本课题通过对乳腺癌细胞切割，截俘细胞轮廓，提取５个细胞核的特征参数及四条判别函数，完成其自动识别的原件系统，５６例标本经该系统处理，正常与癌变的识别率达１００％，导管内癌与浸润性导管癌的识别率为９１％，总识别率９３％，本研究同时提示，复发和转移倾向与乳腺癌细胞核的面积增加有相关性。

In this study,the automatic identifying software system of breast cancer had been completed by dividing the cell of breast cancer,getting contour of the cell, taking out five features of parameters and four identifying functions.Then a morphometric study was performed using the system for histologic specimens of 56 cases. The rate of discernment between normal and malignant reached 100% and between intraductal carcinoma and infiltrating carcinoma reached 91%. Total discerning rate was 93%. This study suggestd that recurrence and metastatic potentiality may correlate with increased nuclear area.