A comparative study of histogram-based thresholding methods for the determination of cell-free layer width in small blood vessels
A comparative study of histogram-based thresholding methods for the determination of cell-free layer width in small blood vessels
复制标题
DOI：
10.1088/0967-3334/31/9/n01
复制
发表时间：
2010-09-01
影响因子：
3.2
通讯作者：
Kim, Sangho
中科院分区：
文献类型：
--
作者：
Namgung, Bumseok;Ong, Peng Kai;Kim, Sangho
我们最近提出了一种基于计算机的方法，利用阈值算法（大津法），以提供一种方便的方式测量无细胞层的宽度在体内和体外。然而，这种方法似乎并不是一种通用的方法，可以应用于所有的微血管研究。因此，我们研究了四种不同的基于直方图的阈值算法（大津，intermode，最小值和第二峰值），以提供一个技术建议，选择一个合适的阈值算法的无细胞层测量。所有的测量都是在用高速摄像机记录的大鼠提睾肌中的微血管流量中进行的。将手动测量的无细胞层的宽度与通过利用四种阈值算法的自动化方法确定的无细胞层的宽度进行比较。实验结果表明，用最小值算法确定的无细胞层宽度与人工测量的无细胞层宽度最吻合。我们得出的结论是，用于测定无细胞层宽度的自动化方法的准确性将取决于图像质量，特别是红细胞核心和背景之间的对比度，这可能由于不同的显微镜设置而不同。因此，在选择最适合实验条件的算法时，可能需要检查几种适当的阈值方法。
We have recently proposed a computer-based method utilizing a thresholding algorithm (the Otsu method) to provide a convenient way of measuring the cell-free layer width in vivo and in vitro. However, this method does not seem to be a universal method that can be applied to all microvascular studies. Thus, we examined four different histogram-based thresholding algorithms (Otsu, intermode, minimum and second peak) to provide a technical suggestion on the selection of a suitable thresholding algorithm for the cell-free layer measurement. All the measurements were taken in microvascular flows in the rat cremaster muscle recorded with a high-speed camera. The width of the cell-free layer manually measured was compared with that determined by the automated method utilizing the four thresholding algorithms. With our experimental system, results showed that the cell-free layer width determined by the minimum algorithm was in best accordance with the manual measurement. We concluded that the accuracy of the automated methods for determination of the cell-free layer width would depend on the image quality, in particular on the contrast between the red blood cell core and background, which might differ due to the different microscopic setup. Therefore, one may need to examine several appropriate thresholding methods when selecting the best suitable algorithm for the experimental conditions.