Exploring highly resolved functional brain connectivity: Current developments and novel perspectives
- Jan 30, 2020
- 2:30 PM - 4:00 PM
- Dr. Britta Pester
- Chair of Computer Graphics and Visualization, Faculty of Computer Science Technische Universität Dresden
- IMB - Seminar
- Main Topic
- The human brain is a complex network consisting of spatially distributed neural assemblies that are connected to each other. In this context, a comprehensive insight into brain processes requires both, the consideration of brain activity as well as an understanding of information flow between and within structures of the underlying neural system (connectivity).
However, the investigation of brain connectivity commonly comes along with an immense volume of data which can be attributed to two reasons:
(1) Spatially high-resolution imaging techniques (such as functional magnetic resonance imaging, fMRI) commonly have a comparably low temporal resolution. This combination of high spatial and low temporal resolution leads to the problem that an adequate connectivity analysis is not feasible by means of conventional methods.
(2) On the other hand, data with a high temporal but low spatial resolution (such as electroencephalography, EEG) allow the application of conventional techniques. However, as soon as time-variant and frequency-selective analyses are used, a huge number of output is produced, leading to serious problems in interpretability.
There are two basic approaches that are commonly used to solve problem (1). One possibility is to restrict the connectivity analysis to a pre-defined subset of several nodes (e.g. fMRI voxel). The other commonly used methodology is to reduce the spatial dimensionality by a suitable coordinate transformation such as principal component analysis and to limit the analysis to a few derived components with the highest variance explanation. Both approaches have the serious drawback that they do not offer an insight into high-dimensional (i.e. highly resolved) connectivity patterns.
Problem (2), the high number of analysis output, is frequently tackled by averaging the results within a certain time and/or frequency interval. However, as a consequence, transient network properties cannot be revealed, and/or frequency-dependent characteristics get lost. Another possibility is to inspect the time-variant and frequency-selective results for a limited number of pre-selected recording sites. This approach does not allow an understanding of the resulting network as a whole and it requires clear hypotheses about relevant brain areas which are oftentimes not available in clinical practice.
In this talk, I will present two approaches that prevent the described problems of the commonly used strategies. To obtain a representation of connectivity for spatially high-dimensional data, a fully multivariate Granger causality approach with embedded dimension reduction is introduced. The challenging handling of the vast number of analysis output is tackled by means of a tensor decomposition of the results offering a complementary view on the derived high-dimensional connectivity patterns. Finally, I am going to introduce current ideas for a novel 3D immersive representation of obtained brain networks in a virtual environment. In this framework, functional EEG networks are considered as fully connected 2D origin-destination data which are transferred to 3D space. Additional attributes (e.g. time, frequency, connectivity threshold) can interactively be adjusted by the user. The goal of this approach is to provide an anatomically intuitive view on the results and furthermore to motivate the user to actively explore the network.
Last modified: Dec 1, 2019, 10:25:06 AM
Institut für Medizinische Informatik und Biometrie (IMB) (Institute for Medical Informatics and Biometry (IMB) Haus 105, Blasewitzer Straße 86 3rd floor, room 3.465)Blasewitzer Straße8601307Dresden
IMBMedizinsche Fakultät Carl Gustav Carus der TU Dresden, Fetscherstr.7401307Dresden
- +49 351 3177 133
- +49 351 3177 225
- Civil Eng., Architecture
- Computer Science
- Electrical and Computer Eng.
- Environmental Sciences
- Linguistics, Literature and Culture
- Mechanical Engineering
- Society, Philosophy, Education