Development of a Quantitative Method of Cancer Cell Diagnosis Based on the Fluorescence Image Analysis of the Paraffin Sections

该项目试图客观地提取癌细胞的病理特征，并开发一种基于核DNA荧光显微图像分析(图像细胞术)的石蜡切片用于癌细胞诊断的定量方法。我们使用了我们开发的、配备了高灵敏度(单色)摄像机的计算机控制的显微图像处理系统。首先，我们用BASIC语言和C语言编制了计算机程序，研究了荧光图像密度的线性定量等基本技术，用碘化丙啶对DNA染色后，对切片中随机切割的细胞体获得的每个核DNA荧光图形进行了图像分析。用图像处理器和个人计算机进行的形态测量产生了许多参数(高达70；大小、形状、染色质)，然后通过多变量、判别和统计分析来处理以阻止…更多的是客观地挖掘癌细胞和正常细胞(上皮细胞、成纤维细胞和淋巴细胞)。然后，我们将其应用于160多例人类胃癌和结肠癌的分析。选择那些具有明显多形性的癌细胞来研究其特征癌细胞。通过这些分析，我们初步选择了癌细胞的特征参数，进而可以在单个细胞的基础上，以0.75-0.93的概率正确地确定显微切片中的癌细胞。我们从硬件和软件两个方面对这一方法进行了总结，以便提出“现场测试模型”的发展，以帮助常规的病理诊断。此外，在现有图像细胞术的基础上，目前正在进行以下方面的研究：1)改进软件(通过建立系统、有效的程序结构和加快计算时间)，2)分析轻微核不典型的癌症病例(如早期癌症、交界性病变等)，3)核不典型程度的定量，4)癌症诊断中的人工智能(AI)技术。较少

The proposed project has attempted to extract objectively pathological features of the cancer cell, and to develop "a quantitative method of cancer cell diagnosis" using paraffin sections based on the microscopic image analysis of the nuclear DNA fluorescence (image cytometry). We used our developed, computer-controlled microscopic image processing system equipped with a high-sensitive (monochromatic) video-camera. At the beginning, we mad e computer programs in BASIC and C languages, and studied basic techniques concerning linear quantitation of the fluorescence image density, etc. After DNA staining with propidium iodide, we carried out image analysis of each nuclear DN A fluorescence figure which was obtained from the randomly cut cell body in the section. The morphometry, performed with both an image processor and a personal computer, yielded many parameters (up to 70; size, shape, chromatin), which were then processed by multivariate, discriminant and statistical analyses to deter … More mine objectively cancer cells and normal cells (epithelial, fibroblastic, and lymphocytic cells). We then applied it to the analysis of more than 160 cases of human gastric and colonic cancers. those cancers being chosen as having the distinctive pleomorphism to study the characteristic cancer cell. Through these analyses, we tentatively selected parameters characteristic for the cancer c ells, and further, could determine correctly cancer cells in the microscopic section in quantitative term with the probability of 0.75-0.93 on a single cell basis. We are summarizing this methodology for both hardware and software in order to propose the development of "the field-test models" in aid of the routine pathological diagnosis. Moreover, studies based on the present image cytometry, are now under way for 1)improving software (by making a systemic, efficient structure of the program and by speed-up of the computation time), 2)analyzing cancer cases with slight nuclear atypism (as occasionally seen in the early cancers, borderline lesions, etc.), 3)the quantitation of the degree of nuclear atypism, and 4)the technique of "artifical intelligence (AI)" in the cancer diagnosis. Less