Data analysis: Texture analysis in high-resolution images

This project is a collaboration of scientists working in two HBP Subprojects:

  • Neuroinformatics Platform (SP5)
  • HPAC Platform (SP7)

Timo Dickscheid (SP2 and SP7) is leading both from a scientific and technical point of view.


The goal of this project is the implementation of a workflow for texture analysis in high-resolution imaging data for the brain atlas that is based on machine learning. It is part of a larger analysis workflow, i.e. the results of this project are intermediate results of the larger workflow and not meant to be provided directly to “end users”. The results must be quality-checked by humans before they can be processed further in the larger workflow. It is planned to realise this manual interaction through a web-based viewer and it can also be used to improve the learning algorithm. When the result has been approved, it should be possible to signal this to the larger workflow so that it continues.

Problems to be solved

The following problems require special attention and need to be solved in co-design with the users for implementing the use case successfully.

  • Provenance tracking of the entire, complex workflow is needed in order to enable reproducibility.
  • Integration of a viewer for HDF5 files into the Collaboratory, that also has access to the data stored in the HPAC Platform infrastructure.
  • The same storage or repository needs to be accessible from the HPC system on which the data is produced, by the viewer and by web server(s) used to transfer the data.