一种基于稀疏协同原型向量重构的鼻咽癌病理细胞协同分类方法

A Synergetic Classification Algorithm of Pathology Cell Images Based on Prototype Vector Fusion with Sparse Decomposition

协同细胞模式分类依据原型向量,特别适用于如细胞分类这种特征不是很明显的分类,而其中原型向量的选取对协同模式的识别结果有着决定性的作用。通过样本的特征变换代替像素点来生成原型向量是研究协同原型向量生成的一种趋势。Contourlet变换是一种新的多尺度几何分析方法,具有很强的方向性和各向异性。采取一种基于contourlet变换的协同原型向量生成方法,通过此种模式融合的而生成框架,应用到鼻咽癌细胞的协同模式分类,同时与传统的协同分类方法和基于curvelet变换及基于contourlet变换的协同分类方法进行比较。结果显示,训练识别率达到96%,测试识别率达到91.33%,生成的初始序参量也更具优势。该方法是一种有效的细胞协同原型向量分类方法。

Synergetic pattern recognition（SPR）,according to prototype vectors,is useful for the cell recognition in which class characteristics are not clear.And the selection of synergetic prototype vectors is critical in the process.There exists a tendency to use character value instead of image pixel for the generation of synergetic vectors.In this paper,the characteristic of contourlet transform was analyzed combined with synergetic pattern recognition.A new fusion method based on contourlet transform for prototype vectors generation was proposed.The coefficients structure and the framework’s fusion procedure were given in details,and the proposed method was tested in nasopharyngeal carcinoma cells classification.Comparisons with other methods,including traditional SPR,the ones based on curvelet transform and based on contourlet transform,were carried out.It was revealed that the recognition rate for training set was 96% and the one testing set was 91.33%.Besides,the proposed method was better in the generation of original order parameters than that of other methods.