Guided Data Cleansing of Large Connectivity Matrices

Information

• Publication Type: Master Thesis
• Workgroup(s)/Project(s):
  • Visualization Group
• Date: January 2019
• Date (Start): 10. January 2018
• Date (End): 29. January 2019
• Open Access: yes
• First Supervisor: Eduard Gröller

Abstract

Understanding the organization principle of the brain and its function is a continuing quest in neuroscience and psychiatry. Thus, understanding how the brain works, how it is functionally, structurally correlated as well as how the genes are expressed within the brain is one of the most important aims in neuroscience. The Biomedical Image Analysis Group at VRVis developed with the Wulf Haubensak Group at the Institute of Molecular Medicine an interactive framework that allows the real time exploration of large brain connectivity networks on multiple scales. The networks, represented as connectivity matrices, can be up to hundreds of gigabytes, and are too large to hold in current machines’ memory. Moreover, these connectivity matrices are redundant and noisy. A cleansing step to threshold noisy connections and group together similar rows and columns can decrease the required size and thus ease the computations in order to mine the matrices. However, the choice of a good threshold and similarity value is not a trivial task. This document presents a visual guided cleansing tool. The sampling is based on random sampling within the anatomical brain hierarchies on a user-defined global hierarchical level and sampling size ratio. This tool will be a step in the connectivity matrices preprocessing pipeline.

Additional Files and Images

Weblinks

• Entry in the publication database of TU-Wien

BibTeX

@mastersthesis{gutekunst_2019,
  title =      "Guided Data Cleansing of Large Connectivity Matrices",
  author =     "Florence   Gutekunst",
  year =       "2019",
  abstract =   "Understanding the organization principle of the brain and
               its function is a continuing quest in neuroscience and
               psychiatry. Thus, understanding how the brain works, how it
               is functionally, structurally correlated as well as how the
               genes are expressed within the brain is one of the most
               important aims in neuroscience. The Biomedical Image
               Analysis Group at VRVis developed with the Wulf Haubensak
               Group at the Institute of Molecular Medicine an interactive
               framework that allows the real time exploration of large
               brain connectivity networks on multiple scales. The
               networks, represented as connectivity matrices, can be up to
               hundreds of  gigabytes, and are too large to hold in current
               machines’ memory. Moreover, these connectivity matrices
               are redundant and noisy. A cleansing step to threshold noisy
               connections and group together similar rows and columns can
               decrease the required size and thus ease the computations in
               order to mine the matrices. However, the choice of a good
               threshold and similarity value is not a trivial task. This
               document presents a visual guided cleansing tool. The
               sampling is based on random sampling within the anatomical
               brain hierarchies on a user-defined global hierarchical
               level and sampling size ratio. This tool will be a step in
               the connectivity matrices preprocessing pipeline. ",
  month =      jan,
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  school =     "Research Unit of Computer Graphics, Institute of Visual
               Computing and Human-Centered Technology, Faculty of
               Informatics, TU Wien",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2019/gutekunst_2019/",
}