Long Time Lapse Microscopy Datasets Computer Science Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Summary: Automated microscopes enable in vivo studies in developmental biology over long periods of time. Time-lapse recordings in three dimensions or higher can produce huge datasets that extend into the terabyte range. However, hardware and software limitations can restrict the amounts of image data a system can cope with. Therefore, to facilitate efficient visualization and quantitative analysis of multi-dimensional microscopy data, we developed TLM-Converter, a tool to reorganize time-lapse microscopy (TLM) data into more manageable portions. Currently, TLM-Converter reads Carl Zeiss confocal laser scanning microscopy (LSM) files and saves the converted output in the Image Cytometry Standard (ICS) or Tagged Image File Format (TIF) file format. The software either fragments oversized files or concatenates multiple files representing single time points. We tested our tool on datasets of live Drosophila specimen recorded using a Zeiss 5 Live inverted laser scanning confocal microscopy system. Aligning the sizes of image files with system resources dramatically enhances the productivity of time-lapse data processing.

Availability and Implementation: The software can be downloaded from ...

The TLM-Converter software is written in C++ with a Java based graphical user interface (GUI) and utilizes the libtiff (http://www.libtiff.org/) and libics (http://libics.sourceforge.net/) libraries. It runs on the Windows platform (XP, Vista and Windows 7) and supports the LSM, ICS and TIF image file formats.

Contact: [email protected]


Time-lapse imaging using automated microscopes can capture developmental processes in live plants and animals at high spatial and temporal resolution [1-4]. Multi-dimensional image acquisition of live specimen performed in 3D over several hours or days and often involving multiple fluorophores can produce huge datasets that extend into the terabyte range. Subsequent analysis of image data involves 3D reconstruction and visualization, segmentation and tracking [5-9]. However, depending on available computational resources and software design, downstream processing of these very large image files can become highly inefficient, if not impossible. For instance, many tools were developed for 32-bit operating systems and are unable to handle data sets of more than 2 Gigabytes. The control software of some automated microscopes allows saving datasets for each time-point individually and then concatenating these files into a single file after recording is completed. This option may encounter the problem that concatenation fails once the data size exceeds a certain limit or that processing of single time point files is rather unproductive. An obvious solution is to fragment large image datasets and to perform analysis in smaller subsets. Since many open source and commercial tools are unable to open subsets of multi-dimensional data, we developed TLM-Converter, a tool to reorganize time-lapse microscopy (TLM). It contains two main modules, one for the conversion and fragmentation of oversized files and another one for the concatenation of multiple stacks representing single time-points.


The TLM-Converter was developed to facilitate the reorganization of time-lapse data using two modules; one for file conversion and fragmentation and one for concatenation (Figure 1).

Conversion & Fragmentation

The main purpose of the conversion and fragmentation module is to divide very large multi-dimensional datasets into user-defined subsets for more efficient downstream image processing. It is useful when limited system resources (e.g. OS, memory) prevent the loading of complete datasets or a required image analysis tool lacks the ability to load subsets of data. Another purpose of fragmentation is to facilitate the parallel processing of time-consuming tasks like deconvolution and 3D segmentation. Input files are Zeiss LSM, ICS and TIF files, while outputs are saved in either ICS or TIF format. ICS serves as the preferred output format as it is an open standard for storing images of any dimensionality and data type that can be read and written by a variety of commercial and open source image processing tools [10]. The software reads and displays the critical parameters of the input files, based on which the user can specify the subset of the data to be stored in the output file in terms of time frames, optical sections and color channels. In batch fragmentation mode, the user can choose the number of time frames to save in each output file.


During recordings in up to 6 dimensions involving multiple locations, colors and time frames, image acquisition software, such as the Zeiss Multiple-Time-Series (MTS) macro, saves time-lapse data as individual Z-stacks. The concatenation module concatenates the individual files into larger user-defined subsets of the time-series experiment that are more convenient for further image analysis. The inputs are .LSM files stored in the same folder whose names contain consecutive numbers that represent the single time points, while the single concatenated output file is saved in the .ICS format. The user needs to select one of the .LSM files and specify a template. The TLM-converter reads the image parameters and counts the files of the time-lapse sequence that match the template. The tool gives the user the flexibility to specify a subset of data to incorporate into the reorganized time-lapse file, including the range of Z-sections and time frames, start time and time intervals, and color-channels.


Figure 1 : (a) GUI of conversion and fragmentation and (b) concatenation module. The user can specify the subset of data to be stored in the output files. In the concatenation module, questions marks in the input filename indicate the time frames.

Image Viewer of ICS files

To check the integrity of very large (>4.3 GB) ICS files, we developed a tool for reading and viewing ICS files. The tool includes a GUI and was implemented in C++ .Net (unpublished).

Image Acquisition and Processing

3D time-lapse imaging of live Drosophila specimen was performed on a Zeiss 5 Live inverted laser scanning microscope. We recorded 4D images of live embryos expressing histone H2AvD-GFP (ref) and 5D images (4D with 2 colors) of prepupae expressing histone-RFP and MHC-tauGFP in muscles [11]. Basic image processing and visualization of the LSM files was carried out using the LSM Image Browser version 4 and the Zen 2009 Light edition (Zeiss). To validate the outputs of the TLM-Converter we used the 64-bit versions of the ImageJ variant Fiji [12], the image deconvolution package Huygens Professional (Scientific Volume Imaging) and the 3D visualization software Imaris X64 v5.7 (Bitplane).


We tested the conversion and fragmentation module of the TLM-Converter with single channel (4D) and dual channel time-lapse image files of up to 18 GB (240 frames) and 69 GB (480 frames), respectively on a Desktop PC with Intel core i7 processor 3.07 GHz and 12GB RAM. The TLM-converter takes approximately 48 minutes to convert 69 GB of dual channel data from LSM into ICS format.

Since operating systems and image processing software impose limits on the size of image files, the single channel 18 GB dataset or in general, very large image files need to be chopped into smaller pieces for further analysis. For instance, in Microsoft Windows environments, 32-bit processes are limited to 2 GB and 4 GB of memory on 32-bit and 64-bit respectively (msdn.microsoft.com). To validate the data integrity of the output files, we used the 64-bit versions of ImageJ, Imaris and Huygens Professional. Imaris and Huygens could read and display ICS output files of up to 4.3 GB, while ImageJ could read files of up to 2.1 GB. When output files exceeded these sizes, image frames located beyond those limits failed to load. Interestingly, these three tools also failed to read frames at the end of LSM files that were longer than 4.3 GB or the maximum number that can be represented by an unsigned 32-bit integer. To confirm that the data of ICS files were intact beyond 4.3 GB, we used an in-house developed software for the display of ICS image files. Following the image fragmentation, parallel processing of computationally intensive routine tasks like deconvolution or 3D segmentation became easier.

To test the concatenation module, we used a time-lapse dual-channel 8-bit dataset consisting of 300 image stacks that each contained 38 optical sections. The software took 11 minutes to concatenate 300 LSM files into a single ICS file of 23 GB. The only other free software that can accomplish this task on a Desktop PC is Zen Light (http://carl-zeiss-microimaging-gmbh.software.informer.com/). However, this tool only allows manual concatenation of one stack at a time until system memory is depleted. Since time-lapse experiments can involve several hundred time frames recorded simultaneously in multiple locations, our concatenation tool removes a bottleneck in image analysis pipelines. Subsequent steps, such as exploring the entire experiment as sequences of maximum intensity projections, can be carried out with higher productivity.

In summary, TLM-Converter is a user-friendly post-processing tool for time-lapse datasets that runs on single desktop computers. Huge multi-dimensional datasets can be reorganized to match available software and hardware resources, saving the need to invest in new computational equipment. Our work illustrates that not all image analysis tools, including various 64-bit version, are fully capable of handling very large image files.