Large-scale measurement of cellular properties

To ensure the validity of the models developed in the Virtual Liver project, they are validated by means of experimental data. Here, the validation of scale-independent models is a particular challenge. The data must be both fine-grained enough to describe the properties of individual cells, and comprehensive enough to represent their distribution across the organ.

  • Computer-based image analysis methods can measure cellular properties in microscopic tissue images automatically and accurately. The imaging of entire organs, however, produces gigabyte-sized tissue images, which are too big to be analyzed with existing methods.
  • Scientists in the Virtual Liver project therefore develop novel analysis methods that can analyze even large tissue images comprehensively and quickly. These methods can recognize different tissue and cell types, for example in order to measure the regeneration activity or the fraction of necrotic tissue in the liver. Another application is the automatic measurement of fatty changes in the liver, which become visible under the microscope as thousands of roundish fat vacuoles (see figure). Fatty changes are an important model parameter in the Virtual Liver Project because they reduce the liver function.
  • In clinical practice, cellular properties are regularly measured in tissue samples in order to assess the aggressiveness of tumors. Since a complete analysis is too time-consuming, it has to be limited to only a few tissue regions. The image analysis methods developed in the Virtual Liver project, for the first time, enable the complete analysis of large tissue regions. In this manner, they do not only provide the data for the validation of scale-independent models, but also have the potential to improve the accuracy of medical diagnostics.
  • Author: André Homeyer

  • Image: Uta Dahmen, Olaf Dirsch and André Homeyer

  • Publications:
    1. Practical quantification of necrosis in histological whole-slide images
    2. A comparison of sampling strategies for histological image analysis
    Comprehensibility: 
    Average: 5 (3 votes)
    (1=bad 5=excellent)