Abstract: This paper introduces an automated classifying method of multi-spectrum bone marrow cell images which combines hierarchical tree and maximum likelihood method, particularly discusses how to confirm the structure of the hierarchical tree, how to select node features and how to construct node classifier. The experimental result shows that the method brought forward in this paper is quite effective for the classification of multi kinds of bone marrow cells, the correct ratio is much higher than simple maximum likelihood method classifier with satisfactory result.

Key words: Hierarchical tree; Maximum likelihood method; Multi-spectrum; Pattern recognition; Classifier; Bone marrow cell

摘要： 介绍了一种结合决策树和极大似然法的多光谱骨髓细胞图像自动分类的方法，详细讨论了怎样确定决策树的结构，应用散点图挑选节点特征以及构造节点分类器。给出了分类结果。实验结果表明，文章提出的基于决策树的细胞分类方法对于多类的骨髓细胞分类识别极为有效，分类准确率远远高于单纯的极大似然法分类器，且分类的速度也令人满意。

关键词: 决策树；极大似然法；多光谱；模式识别；分类器；骨髓细胞