论文部分内容阅读
应用计算机辅助形态定量法探讨胃癌脱落细胞的图象分析与识别。实验将胃刷涂片中的腺上皮细胞分成三类:(1)正常;(2)核异质;(3)癌。实验通过细胞图象增强,图象分割,抽取细胞的六维特征,即(1)细胞面积;(2)核面积;(3)核浆比;(4)核的形状因子;(5)核的平均灰度;(6)核内光密度的均方差。最后设计和训练分类器,从而实现了一套计算机辅助形态定量分析系统。该系统分别对样本模式、样本模式与未知样本随机混合以及未知样本重复检验,其分类符合率前二者均为98%,后者为95%。分类器的性能取得令人满意的结果。
Application of Computer Aided Quantitative Assay for Image Analysis and Identification of Exfoliated Cells in Gastric Cancer . The experiment divided the glandular epithelial cells in the stomach brush smear into three categories: (1) normal; (2) nuclear heterogeneity; (3) cancer. The experiment was performed by image enhancement and image segmentation to extract the six-dimensional characteristics of the cells, namely (1) cell area; (2) nuclear area; (3) nuclear-to-plasma ratio; (4) nuclear shape factor; (5) nuclear The average grayscale; (6) The mean square deviation of optical density within the nucleus. Finally, the classifier was designed and trained to realize a computer-assisted morphometric quantitative analysis system. The system separately tests the sample pattern, the sample pattern and the random sample of the unknown sample, and repeats the unknown sample. The classification coincidence rate is 98% and the latter is 95%. The classifier performance achieved satisfactory results.