【摘要】 采用共焦拉曼光谱仪,测量了正常乳腺组织和浸润性导管癌组织的阵列显微拉曼谱3 000多个,利用自编软件对谱数据进行K阶簇分析(K-means cluster analysis),得到各簇的平均谱以及各谱的簇类别。各簇平均谱与商售的甘油三酸脂、胶原和肌动蛋白等纯生化物质的拉曼谱具有很高的相关性,说明K阶簇分析能够区分细胞质、细胞间质及脂肪等不同组织形态的拉曼谱;依据各谱的簇类别构建阵列拉曼谱的簇分析成像能够显现乳腺上皮组织的生化成分分布及乳腺管结构,与乳腺组织形态模型成像结果相互印证。本文研究预示簇分析是研究乳腺组织拉曼谱及提取诊断信息的有效手段,优势在于不需要知道组织样品的生化信息。更多还原

【Abstract】 More than 3 000 Raman spectra are collected on the samples of the normal human breast duct epithelia and the infiltrating duct carcinoma tissues by means of confocal Raman spectroscopy using a microscopic mapping approach with the sample volume of ~2 μm3.Using K-means cluster analysis,these spectra are classified with a program in Matlab 6.5,so the average spectra for all classes are calculated and the class indices for all spectra are gotten.These average spectra have respectively high correlation coefficients with the ones of the commercially available chemicals triolein,collagen(type I),actin etc,which construct the main substance of lipid,cell cytoplasm,and extra-cellular matrix etc.The mapping images,created with the class indices of the mapping spectra,show the bio-chemicals′ distribution and the structure of human breast duct and they are very similar to the images made by means of the morphological modeling.These results indicate that the K-means cluster analysis is a helpful method to understand the Raman spectra of human breast tissues and to collect diagnostic information with the Raman spectra for breast tumor.The Raman images can be created by K-means cluster analysis without knowing the spectra of the bio-chemicals in breast tissue samples.更多还原