Weinreb M, Zajicek G, Levij I S
Anal Quant Cytol. 1984 Sep;6(3):206-10.
The nuclei of 21 normal and 23 malignant epithelial cells from gastric smears obtained by brushing were analyzed by a black-and-white video camera under computer control. Each nucleus was digitized and its relief smoothed and displayed. A Sobel operator determined the nuclear boundaries and nuclear core area. Eighteen nuclear parameters (form descriptors and gray-value descriptors) were extracted for each nucleus and used as variables for discriminant analysis. Nine of these parameters proved useful for discrimination between normal and malignant gastric epithelial cells, with a correct classification rate of 100%. Of these, the best discriminating variables were the maximal gray value in the core, the maximal horizontal and vertical diameters, the core area and the mean derivative value in the core.
通过计算机控制的黑白摄像机对21例正常胃涂片和23例刷取法获得的恶性胃上皮细胞涂片的细胞核进行分析。每个细胞核进行数字化处理，其轮廓平滑后显示。使用索贝尔算子确定细胞核边界和核仁区域。为每个细胞核提取18个核参数（形态描述符和灰度值描述符），并用作判别分析的变量。其中9个参数被证明对区分正常和恶性胃上皮细胞有用，正确分类率为100%。其中，最佳判别变量为核仁中的最大灰度值、最大水平和垂直直径、核仁面积以及核仁中的平均微分值。
