白细胞的自动识别和分类

白细胞的分类目前已成为医院的常规检查项目之一,检查结果对于临床诊断有着十分重要的意义。然而,显微镜下的这项工作是十分耗时的,且不同的检验人员有着不同的分类误差。为此,本文初步探讨了用图象处理和模式识别的方法对白细胞进行自动识别和分类,希望能以计算机来帮助人的眼脑系统,从而减轻检验人员的工作强度,提高分类的效率和一致性。本文运用了图象处理手段,用求熵阈值法对细胞图象进行分割,提取细胞及细胞核;用全向跟踪法提取细胞轮廓和核轮廓,同时,提出了一种快速的区域生长算法对轮廓内区域作填充。本文对每幅细胞图象提取了一个五维的特征向量,采用多两类分类和留一法,对有限的样本作训练,以设计Bayes分类器,对白细胞逐层分类,并对分类效果作出检验。本文通过以上方法的应用,白细胞的分类取得了比较好的结果。

The Recognition and Classification of White Blood Cells

The exploration of automatio recognition and classification of white blood cells by the method of image processing and pattern recognition is presented in this paper Image processing is used to obtain the entropy threshold, which is the most important in image segmentation to isolate the cell and nucleus of cell from background. The method of tracing in all direction is used to extract the edges of cell and necleus. An algorithm of quick area expand is put forward to fill the area in edge. In this paper, a feature vector with fivedimensions is to be obtained from each cell image and the limited samples are trained to be designed to the Bayes classifier. The white blood cells are then classified level by level and the classification error is estimated.