Zuse Institute Berlin, Department of Visualization and Data Analysis, Takustrasse 7, 14195 Berlin, Germany.
J Struct Biol. 2012 May;178(2):129-38. doi: 10.1016/j.jsb.2011.12.004. Epub 2011 Dec 13.
The ability to rapidly assess microtubule number in 3D image stacks from electron tomograms is essential for collecting statistically meaningful data sets. Here we implement microtubule tracing using 3D template matching. We evaluate our results by comparing the automatically traced centerlines to manual tracings in a large number of electron tomograms of the centrosome of the early Caenorhabditis elegans embryo. Furthermore, we give a qualitative description of the tracing results for three other types of samples. For dual-axis tomograms, the automatic tracing yields 4% false negatives and 8% false positives on average. For single-axis tomograms, the accuracy of tracing is lower (16% false negatives and 14% false positives) due to the missing wedge in electron tomography. We also implemented an editor specifically designed for correcting the automatic tracing. Besides, this editor can be used for annotating microtubules. The automatic tracing together with a manual correction significantly reduces the amount of manual labor for tracing microtubule centerlines so that large-scale analysis of microtubule network properties becomes feasible.
快速评估电子断层扫描 3D 图像堆栈中微管数量的能力对于收集具有统计学意义的数据集至关重要。在这里，我们使用 3D 模板匹配来实现微管追踪。我们通过将自动追踪的中心线与大量早期秀丽隐杆线虫胚胎中心体的电子断层扫描的手动追踪进行比较，来评估我们的结果。此外，我们还对另外三种类型的样本的追踪结果进行了定性描述。对于双轴断层扫描，自动追踪平均产生 4%的假阴性和 8%的假阳性。对于单轴断层扫描，由于电子断层扫描中的缺失楔形，追踪的准确性较低（16%的假阴性和 14%的假阳性）。我们还实现了一个专门用于纠正自动追踪的编辑器。此外，这个编辑器还可以用于标注微管。自动追踪和手动校正显著减少了手动追踪微管中心线的工作量，使得大规模分析微管网络特性成为可能。
