摘要
将图像纹理特征和物质含量测量参数—光密度值等结合起来,组成一种新的特征向量,用于分析肝癌、肝硬化时淋巴细胞化学染色核图像的微小变化。通过改进血液涂片载体,以利于提高光密度值的精度和核纹理分析。采用小波变换结合经验模式分解获取图像表面纹理特征,再与多种图像光密度值和图像能量等结合,构成一种图像组合特征多维向量,用支持向量机完成分类识别。实验结果表明,该法较好地区别了正常人与肝硬化、肝癌病人以及肝硬化与肝癌病人的外周血淋巴细胞。分析了淋巴细胞核中核酸、蛋白质等与核表面纹理及化学着色之间的关系。解决了淋巴细胞核染色表面微小变化人眼不能定量区分识别的困难。图像组合特征多维向量可以有效地反映不同淋巴细胞核表面间的纹理差异,可作为识别淋巴细胞核染色表面变化新的特征量。
An eigenvector,composed by the image texture features,the optical density and image energy,was applied to analyze slight changes after lymphocyte staining on lymphocyte nucleus images of liver cancer and hepatocirrhosis patients.Improved slide of blood smear was used to enhance the precision of optical density and nuclear texture analysis.The wavelet transformation and the empirical mode decomposition were combined to obtain the image surface texture feature.The texture features,the image optical density and the energy were composed as a multi-dimensional vector,and the pattern recognition was performed with support vector machines.The experimental result showed the proposed method could distinguish normal and hepatocirrhosis,liver cancer patient′s lymphocyte with high accuracy.It was demonstrated that image feature multi-dimensional vectors were able to differentiate with different type of lymphocyte nucleus texture surfaces efficiently,and were promising in the recognition of lymphocyte changes in the nucleus of lymphoeytes staining as new features.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2010年第2期190-194,共5页
Chinese Journal of Biomedical Engineering
基金
浙江省教育厅科研项目(20070930)
关键词
图像处理
小波变换
淋巴细胞
核酸
image processing
wavelet transformation
lymphocyte
nucleic acid
