红细胞SEM图像的三维重构及形状特征提取

Research on three-dimensional reconstruction and shape feature extraction of red blood cell SEM image

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%。

Many diseases are related to the changes of erythrocyte morphology in clinical experience and diagnosis. Therefore the morphological characteristics of red blood cells can be analyzed in the diagnosis of the disease. The template matching method is used to find the location of the twelve types of morphological variation the red blood cell image,PCA and LDA algorithm is used to achieve the feature selection and extraction of twelve kinds of red blood cells. The algorithm is improved for the light of noise problem. Through the experimental data,the data are compared for some cells of some classification difficult,some cells of feature are extracted and classified. Experiments shows that the algorithm and the improved method can effectively differentiate and extract different types of red blood cells. The accuracy of classification is 92. 7%.