【作者】 宁旭；

【导师】 谢利利；

【作者基本信息】 重庆大学 ， 光学工程， 2002， 硕士

【摘要】 目前医学病理诊断主要是由医务人员对细胞组织切片在显微镜下进行形态学观察，并依据经验得出诊断结论。这种方式定性的成分居多，客观性不足，在诊断科学逐步走向量化的道路上，有必要使用计算机信息技术推动病理诊断的自动化、科学化。本文针对上述问题，用计算机图像处理及模式识别等信息技术对显微细胞图像的自动分析和分类的方法进行了研究，并针对医学图像分析中的难点（例如，显微切片图像背景复杂，分割困难），提出了基于归一化彩色空间和RGB，HSV彩色模型的两类分割方法：①利用模式识别技术中关于特征向量空间聚类的方法实施真彩色分割。②利用HSV模型，采用最大类间方差的阈值分割。这两种方式有效地利用了多维特征空间对于分割目标所提供的信息，使分割的准确性有了较大的提高。论文还进一步探讨了有关细胞形态及色度测量的技术，在原有的技术基础上完善了目标识别及轮廓跟踪的算法，使之可以测定多种参数，并提出了多项衡量细胞特征的指标，对细胞分类具有重要的指导意义。在实验的基础上，给出了区分正常细胞与肿瘤细胞的有效指标，并对细胞分类的方法作了探讨。全文给出了解决细胞显微图像的自动化检测的技术途径和实现方案，为进一步的研究打下了基础。更多还原

【Abstract】 At present, medical workers mainly make pathological diagnosis through morphological observing to the section of cell tissue under microscope, and draw conclusion by experience. This kind of method is mainly qualitative and lack of objectivity. With the development of the technology, it is necessary to use the computer information technology to promote the pathological diagnosis to be more automatic and scientific.In view of the above mentioned problem, the author adopts information technology such as image processing and pattern recognition to research into the method of automatic analysis and classification. In accordance with the difficulty in medical image analysis (for example, the background of microimage of section is complicated and is difficult to be segmented.), the paper puts forward two kinds of segmentation methods based on standardized colorful space and RGB and HSV colorful model. First is the true-color segmentation using the method about eigenvector space cluster in pattern recognition. Second is the threshold segmentation adopting the biggest class separation variance and HSV model. This two kinds of methods make a good use of the information supplied by multiple-dimensional feature space and enhance the accuracy of segmentation. The paper further discusses the technique about measuring the shape and chroma of cell. On the basis of the existed technology, the paper perfects the arithmetic of target recognition and contour tracking and enables it to measure several kinds of parameters. The paper also brings forward several indexes to measure cells, which is of instructive meaning to cell segmentation. On the basis of experiments, the author presents a effective index to distinguish abnormal cells from normal ones and examines the method of cell classification. Above all, the paper puts forward a way of automatic detection of microimage of cell and lays a good foundation for a further study.更多还原