What is Thor?
THOR Center for Neuroinformatics
  overview     human brain project consortium  

The consortium supports three closely connected research projects and four "cores", the central theme of the work is modeling and visualization of spatial and temporal patterns of functional activation in the living human brain as imaged by [15O]water PET and fMRI.

The consortium consists of two informatics components - Project 3 (Interactive visualisation for Neuroimaging) and core D (Neuroinformatics) and two brain/behavioral components - Project 1 (Temporal Resolution of fMRI) and Project 2 (Reproducible Features of Functional Neuroimages); Cores B and C contain well-defined research components - 3D brain warping (Core B) and behavioral correlates of functional neuroimaging data (Core C).

It consolidates a earlier collaborative research program under the Human Brain Project and involves neuroscientists, neuropsychologists, physicists, mathematicians, statisticians, computer scientists and health informatics specialists at five U.S. and four institutions outside the U.S.:

  • Minneapolis VA Medical Center (USA) University of Minnesota (USA).
  • Florida State University (USA).
  • Massachusetts General Hospital (Harvard Medical School, USA).
  • McLean Hospital (Department of Psychiatry, Harvard Medical School, USA).
  • Akita Research Institute of Brain and Blood Vessels (Japan).
  • National University Hospital, Rigshospitalet (Denmark).
  • Technical University of Denmark.
  • Niels Bohr Institute (Denmark).
The consortium brings together two academic MRI/fMRI centers and three academic PET centers, hence, provides a unique opportunity to study the spatial and temporal patterns in functional neuroimages and to compare and contrast neuroimages acquired during rigorously standardized neurobehavioral protocols using 1.5T fMRI, 4T fMRI, and [15O]water PET .

The main activity for the DTU-group lies in the following projects and cores:

Project 2
Reproducible Features of Functional Neuroimages (S.C. Strother, PI; Lars K. Hansen, Co-PI) has as its principal goal the identification of reproducible features in PET and fMRI datasets - within and between participating sites and across imaging modalities. This project will also assess 1) the impact of non-linear 3D transformations ("warps") for intersubject registration and removal of geometric distortion from 4T echo-planar fMRI images and 2) oriental-occidental anatomical and functional differences on the reproducible features identified using two data analysis strategies - linear models (e.g., MANOVA, ANCOVA, PCA), and non-linear models (e.g., artificial neural networks, k-nearest neighbors). In addition, Project 2 investigators will develop and, in collaboration with Project 3 and the informatics core (Core D) make available on the WWW benchmark functional activation datasets, analysis results and associated modeling software. PET and fMRI datasets will be provided by Project 1 and core B.
Project 3
Interactive Visualization for Neuroimaging (D.A. Rottenberg, PI) will develop an interactive visualization environment for exploring multidimensional functional data fused with a 3D structural image volume. This environment will integrate a suite of existing multidimensional display tools (surface and volume rendering, slice galleries and projections) with a rich set of navigational tools, pointers to atlas information and access to the HBP Data Repository (supported by Core D). Furthermore, Project 3 will develop a VRML environment that supports the presentation of modeling and data-analytic results ("case studies") in an interactive form that is shareable locally and over the Internet. VRML translations will be initiated from the interactive visualization environment and will be designed to minimize the need for manual editing prior to export.
Core B
will acquire, archive and ensure quality control of structural and functional scans at the participating fMRI and PET centers. In addition, Core B investigators will implement two non-linear 3D warp algorithms for intersubject registration and evaluate "intersubject subspace variability" and "model prediction error" as metrics for optimal warp tuning and warp selection.

© Copyright 1998, THOR Center for Neuroinformatics
Page maintained by Jan Larsen