Principles of cell and tissue organization
MPI-CBG MPG
 


Projects


Imaging and quantitative multi-parametric image analysis (QMPIA)

MotionTracking Software

Quantitative image analysis is built on the MotionTracking platform. MotionTracking provides a set of visualization and image manipulation tools as well as a rich set of mathematical algorithms for the general data analysis and image processing. The MotionTracking project was initiated in 2001 to track the movement of intracellular vesicles. Later it was expanded to be used for high-performance (supercomputer) calculations of high-throughput genome-wide screens data. MotionTracking is the implementation of a “language-stack” paradigm, where semantically different parts of applications are implemented on the basis of different programming languages. The general logic flow of data processing, high-level logic of algorithms and user interface is implemented on the dynamic object oriented language Pluk. The number crunching algorithms are implemented on C++ and OpenCL. The interface of the operating system GUI and object-oriented database access are implemented by specialized declarative languages (OWML and ODML). This approach allows automating the code generation and accelerating the prototyping and logic design of algorithms. Currently, MotionTracking includes more than 100 types of statistic information that could be extracted from the microscopy images.

MotionTracking includes a set of algorithms for:

  • The detection of pleomorphic objects with a size ranging from the diffraction limit to few micrometres and intensities varying in 2-3 orders of magnitude with sub-pixel accuracy 
Fig. 1

The detection of pleomorphic objects with a size ranging from the diffraction limit to few micrometres and intensities varying in 2-3 orders of magnitude with sub-pixel accuracy

GFP-Rab5a labelled endosomes in A431 cells. Pleomorphic objects with size from diffraction limit to few micrometres (endosomes) are fitted by sum of base functions: 

  • The tracking of intracellular objects

GFP-Rab5a labelled endosomes in A431 cells. Averaged over 15 minutes movement of endosomes in the cell presented as a flux direction.

  • The reconstruction of PALM/STORM image 

Super-resolution image reconstruction (PALM/STORM). Left panel – conventional confocal microscopy image. Right panel – super-resolution image reconstruction. HeLa cells, red color is EGF, green color transferrin.

  • The analysis of object-oriented intracellular marker colocalization 

The analysis of object-oriented intracellular marker colocalization

Object-based colocalization answers biologically relevant questions:

  1. Which percentage of vesicles carries two/three/four markers?
  2. Which percent of total marker reside on two/three/four marker carrying endosomes? 
  • The analysis of 2D nuclei and cells identification and morphology

2D identification of cells and nuclei in case of low contrast cell borders. HeLa cells; Blue – DAPI staining, green – transferrin, red – EGF.

  • The detection of thin cellular processes (neurites) 

Identification of thin process with low intensity and contrast (neurites).

  • 3D nuclei cells and tubular networks identification in thick tissue slices

3D nuclei cell and tubular network identification in thick tissue slices (mouse liver). Red - endothelia network, green bile canaliculi, hepatocytes are marked by different colours for better visibility.

Development of image analysis for the 3D-reconstruction of liver tissue

The morphological characterization of liver tissue at high resolution required the design of a new set of approaches to perform imaging, image processing and analysis. To this end, we developed a range of new algorithms and integrated them in the MotionTracking software. Collectively, these algorithms allows us to do the following:

  1. Align z-stacks acquired in different modalities (2-photon and 1-photon) and correct standard errors of the scanning confocal microscope;
  2. Align the image stacks coming from adjacent regions of large tissue portions;
  3. Achieve the reliable, high-resolution segmentation of liver structures (e.g. sinusoidal network, endothelial network, hepatic cell walls, etc);
  4. Provide proper statistical description of segmented structures.

We will continue the development of image processing and analysis algorithms to achieve a comprehensive and accurate digital representation of the 3D organization of liver tissue. This will serve to develop a multi-scale mathematical model of liver fluid mechanics based on high resolution imaging and quantitative image analysis of liver tissue. 

Digital representation of tubular networks in liver tissue

3D reconstruction and skeletonization of sinusoidal and bile canaliculi networks from confocal microscopy images at high-resolution of thick slices of mouse liver. Nuclei in blue, sinusoidal and bile canaliculi networks in red and green respectively.

Zerial lab
funding