论文部分内容阅读
临床经验及诊断中表明许多疾病都与红细胞形变有关。因此分析红细胞的形态特征可以辅助诊断病人的病情。运用模板匹配方法寻找12类形态变异的红细胞子图像的位置,应用PCA和LDA算法对12类产生形变的红细胞进行特征选择和提取,并针对噪声问题对算法进行了改进。通过实验数据对一些分类困难的形变细胞做进一步的数据对比及特征提取分类。实验表明,该算法及改进的方法能有效区分并提取出不同类型的红细胞,分类的准确率达到了92.7%。
Clinical experience and diagnosis show that many diseases are associated with red cell degeneration. Therefore, the analysis of the morphological characteristics of red blood cells can help diagnose the patient’s condition. The template matching method was used to find the location of 12 types of red blood cell sub-images with morphological variations. PCA and LDA algorithms were used to select and extract 12 kinds of deformed erythrocytes, and the algorithm was improved according to the noise problem. Through experimental data for some of the difficult classification of deformed cells to do further data comparison and feature extraction classification. Experiments show that the algorithm and the improved method can effectively distinguish and extract different types of red blood cells, the classification accuracy rate reached 92.7%.