线粒体形态学全自动定量分析方法

Automatic quantitative analysis on mitochondrial morphology method

为快速准确地对线粒体形态进行定量分析,提出了一种基于多方向模板响应和追踪的定量方法。该方法先从原始图像中提取种子点,建立16对方向模板。通过模板响应矩阵确定准中心点处的线粒体边界及中心位置。然后从种子点出发,利用边界和中心点进行追踪,得到线粒体的个数、平均长度、平均面积等特征参数。实验结果表明:正常组和游离胆固醇过载的病变组平滑肌细胞内线粒体形态有显著性差异(P<0.05),该方法较现有半自动定量方法处理速度更快,准确率更高。

In order to achieve fast and accurate quantification on the feature of mitochondrial structure, a new method based on multi-directional template response and tracing algorithm was presented. Firstly, the seed points were selected from the original mitochondrial images. Edge points and centerline points of mitochondria structure were detected using directional template response on 16 directions. Then, it initiated from the seed points, and automatically traced the mitochondrial structure recursively by detecting the edge points. The number, average length and average area of mitochondria were calculated from the tracing process. The experiments show that mitochondrial morphology is significantly different （P0.05） between controls and smooth muscle cells with overloaded free cholesterol. The proposed approach is more efficient and more accurate than the interactive method.