Live-Cell Imaging and Automation of Image Analysis Group, Imaging Informatics Division, Bioinformatics Institute (BII), Agency for Science, Technology and Research (A*STAR), Singapore.
BMC Bioinformatics. 2012;13 Suppl 17(Suppl 17):S14. doi: 10.1186/1471-2105-13-S17-S14. Epub 2012 Dec 13.
Metamorphosis in insects transforms the larval into an adult body plan and comprises the destruction and remodeling of larval and the generation of adult tissues. The remodeling of larval into adult muscles promises to be a genetic model for human atrophy since it is associated with dramatic alteration in cell size. Furthermore, muscle development is amenable to 3D in vivo microscopy at high cellular resolution. However, multi-dimensional image acquisition leads to sizeable amounts of data that demand novel approaches in image processing and analysis.
To handle, visualize and quantify time-lapse datasets recorded in multiple locations, we designed a workflow comprising three major modules. First, the previously introduced TLM-converter concatenates stacks of single time-points. The second module, TLM-2D-Explorer, creates maximum intensity projections for rapid inspection and allows the temporal alignment of multiple datasets. The transition between prepupal and pupal stage serves as reference point to compare datasets of different genotypes or treatments. We demonstrate how the temporal alignment can reveal novel insights into the east gene which is involved in muscle remodeling. The third module, TLM-3D-Segmenter, performs semi-automated segmentation of selected muscle fibers over multiple frames. 3D image segmentation consists of 3 stages. First, the user places a seed into a muscle of a key frame and performs surface detection based on level-set evolution. Second, the surface is propagated to subsequent frames. Third, automated segmentation detects nuclei inside the muscle fiber. The detected surfaces can be used to visualize and quantify the dynamics of cellular remodeling. To estimate the accuracy of our segmentation method, we performed a comparison with a manually created ground truth. Key and predicted frames achieved a performance of 84% and 80%, respectively.
We describe an analysis pipeline for the efficient handling and analysis of time-series microscopy data that enhances productivity and facilitates the phenotypic characterization of genetic perturbations. Our methodology can easily be scaled up for genome-wide genetic screens using readily available resources for RNAi based gene silencing in Drosophila and other animal models.
昆虫的变态将幼虫转化为成虫的身体结构，包括幼虫的破坏和重塑以及成虫组织的产生。幼虫向成虫肌肉的重塑有望成为人类萎缩的遗传模型，因为它与细胞大小的剧烈变化有关。此外，肌肉发育适合在高细胞分辨率下进行 3D 体内显微镜检查。然而，多维图像采集会导致大量数据，这需要在图像处理和分析方面采用新方法。
为了处理、可视化和量化在多个位置记录的时间 lapse 数据集，我们设计了一个包含三个主要模块的工作流程。首先，之前介绍的 TLM-converter 连接了单个时间点的堆栈。第二个模块，TLM-2D-Explorer，创建最大强度投影以快速检查，并允许多个数据集的时间对齐。预蛹期和蛹期之间的过渡作为参考点，用于比较不同基因型或处理的数据集。我们展示了如何通过时间对齐揭示参与肌肉重塑的 east 基因的新见解。第三个模块，TLM-3D-Segmenter，对多个帧上的选定肌肉纤维进行半自动分割。3D 图像分割由 3 个阶段组成。首先，用户将种子放置在关键帧中的一个肌肉中，并基于水平集演化执行表面检测。其次，表面传播到后续帧。第三，自动分割检测肌肉纤维内的核。检测到的表面可用于可视化和量化细胞重塑的动态。为了估计我们的分割方法的准确性，我们进行了与手动创建的真实情况的比较。关键帧和预测帧的性能分别达到 84%和 80%。
我们描述了一种用于高效处理和分析时间序列显微镜数据的分析管道，该管道提高了工作效率，并促进了遗传扰动的表型特征描述。我们的方法可以轻松扩展到使用基于 RNAi 的基因沉默在果蝇和其他动物模型中进行全基因组遗传筛选的情况下，使用现成的资源。
