Images


DTU Informatics >>> HBP > Image

VRML image of brain
    activations Corner cube enviroment image of brain activation Corner cube enviroment image of brain activation

VRML screen shots showing a number of different studies: Focal brain activations obtained with statistical analysis of images from a PET scanner (Positron emission tomography scanner). Some of the studies were performed by Ian Law, others were extracted from the BrainMap database.

The surface of the cortex was made by Heather Drury.

The visualizations are also available as 3-dimensional models from the VRML and Projects web-pages, and similar models can be constructed with the Brede Matlab toolbox.

Finn Årup Nielsen; 1998 Feb.


Anaglyph image

Anaglyph stereoscopic image

[ 960x725 JPG 90kB ] Amber/blue anaglyph stereoscopic image of the cortex. A pair of glass is required to view this properly: Left eye glasses should be amber and right eye should be blue. Such glasses are distributed in connection with the 3D Mania - Encounter In The Third Dimension movie [1] that uses the so-called ColorCode 3-D technique [2].

The image is generated with the Brede Matlab toolbox. The target is set at (0, -0.025, 0) and the left eye cameraposition (viewpoint) is at (-0.36 -0.176, 0.0052) meters with a viewangle at 20. When viewed on a my screen the posterior superior temporal gyrus (the center of the image) appears approximately 5 centimeters outside the screen. The object displayed is the cortex surface from the Visible Human atlas generated by Heather A. Drury [3].
  1. 3D Mania - Encounter In The Third Dimension, nWave Pictures, www.nwave.com, Brussels, Belgium.
  2. ColorCode 3-D ApS, Lyngby, Denmark
  3. David C. Van Essen, Heather A. Drury (1997), Structural and Functional Analyses of Human Cerebral Cortex Using a Surface-Based Atlas, Journal of Neuroscience, 17: 7079-7102.
Finn Årup Nielsen, 2001 June.

Saliency map Saliency map Saliency map Saliency map

Images showing a saliency map from four different views. The saliency map [1] is a representation of the importance of brain regions for a neural network model of an activation study. In an activation study subjects are either active (say, doing targetted saccadic eye movements) or resting in a baseline reference state. The present saliency map corresponds to a saccade eye movement study carried out by Dr. Ian Law at the PET-center at the National University Hospital of Copenhagen.

Volume rendering can be used for visualizing saliency maps. The images below are produced by volren, and show a saliency map overlaid an MRI image - the data set is the same eye-movement experiment as above, the visual cortex is clearly activated.

Saliency map - volume rendering Saliency map - volume rendering Saliency map - volume rendering

With volume rendering, the volume can be made partly transparent, - in the pictures below the anatomical image has been made transparent so that we can see the saliency map inside.

Saliency map - volume rendering Saliency map - volume rendering Saliency map - volume rendering

  1. Niels Mørch, Ulrik Kjems, Lars Kai Hansen, Claus Svarer, Ian Law, Benny Lautrup, Steve Strother, Kelly Rehm, Visualization of Neural Networks using Saliency Maps [ Postscript ]

Niels Mørch; 1995.


Reconstructed PET image Reconstructed PET image Reconstructed PET image Reconstructed PET image

Four images showing stacked 2D reconstructed PET images. The colors only show reflections in surface of the brain. The images are generated from a reconstruction (filtered backprojection) of projections [1].

  1. Peter A. Toft, Iterative Methods for Reconstructing PET Images, 1995 [ PostScript ]

Peter Toft, 1995.


Stack of images with reconstructed PET images

A 2D reconstructed PET image showing a horizontally sliced section of the brain. The image is generated from a reconstruction (filtered backprojection) of projections [1].

  1. Peter A. Toft, Iterative Methods for Reconstructing PET Images, 1995 [ PostScript ]

Peter Toft,1995.


MRI image

To reduce variance in functional group studies we morph subjects anatomically. In this demonstration the MRI scans of two subjects are seen (right and left). The subject on the left side is then morphed to fit the subject to the right, as seen in the center display. The morphing is truely 3D, and based on local match.

Ulrik Kjems, Peter Philipsen, 1995


Brain metabolism

The two images show a map of the glucose metabolism in a slice of the brain estimated from a set of dynamic PET-scans and a blood input curve. In the left image the rate-constants in Sokoloff's differential equations are fitted direct at a pixel by pixel basis, and from these rate-constants the glucose metabolism is calculated. The right image shows the results of using a feed-forward neural network to estimate the glucose metabolism. The neural network methods helps in filtering the images and reducing the computational burden of the estimation process. In an article [1] presented at NNSP'95 the principle is described in detail.

  1. Claus Svarer, Ian Law, Søren Holm, Niels Mørch, Olaf Paulson, Estimation of the Glucose Metabolism from Dynamic PET-scans using Neural Networks, NNSP'95 [ Compressed PostScript ]

Claus Svarer, 1995


VRML screen shot of result of brain image analysis VRML screen shot of
      result of brain image analysis VRML screen
      shot of result of brain image analysis VRML screen shot of
      result of brain image analysis

VRML-screen shots show the result from a neural network analysis of an fMRI study: left hand sequential finger opposition. The functional blobs are put into the Talairach coordinate system. Orange blobs are first term saliency, yellow are second term saliency. The magneta blobs are from an important principal component, and the red blobs are "subtracted baselines" found through a "backprojection" from the hidden units in the neural network. Also shown are the Penfield Homunculus somatosensoric (green textured plane) and motoric map (blue textured plane), Brodmann numbers, and a "chessboard" representing the saliency for the 17th slice. The orthographic image to the left has a surface for the full activation scan, and the right perspective image has a plot showing the integrated blob activity as a function of time.

Finn Årup Nielsen, 1996.


Image of MRI
      brain tissue segmentation Image of MRI
      brain tissue segmentation

Pictures from Brain matter classification - Allan Rene Rasmussen's Master Thesis: Voxels in T1 and T2 weighted MRI images are by a human expert classified into four classes: white matter (neuron connections), grey matter (neurons), CSF (brain fluid) and lesions. An artificial neural network is trained to imitate this classification.

The pictures show three classes: the grey matter as pink, the white matter as yellow-white and the CSF as green-red.

Allan Rene Rasmussen, Finn Årup Nielsen, 1996.


Images of cortex Images of cortex

Screenshots from a VRML display of a "mouth" study with Heather Drury's and David Van Essen's triangulation of the right hemisphere of the Visible Human dataset. The mouth activations are the yellow areas (blobs). The small red areas in the second image are activations from a study of S. E. Petersen "Passive Words" extracted from the BrainMap database.

The Visible Human surface has been moved 6mm in the anterior direction compared to the original Drury coordinates.

Van Essen Lab with the Visible Man surface-based atlas: http://v1.wustl.edu/

Finn Årup Nielsen, 1998 January


Information space

Information space

Information visualization or process visualization

Screenshot of a VRML-world visulization of the process in a functional neuroimaging study: From the initial idea and hypothesis, through protocol design, scanning, preprocessing, and statistical analysis to visualization and the final conclusion.

A similar torus world is available in VRML from the paper Visualizing data mining results with the Brede tools.

Finn Årup Nielsen, 1996–1998, 2009


Brain image HBP THOR Center for Neuroinformatics, Human Brain Project Repository (This server)
THOR THOR Center for Neuroinformatics
ISP Digital Signal Processing
IMM Informatics and Mathematical Modelling
DTU Technical University of Denmark

© DTU, 1995-2002, 2005, 2006, 2009
hendrix.imm.dtu.dk/image/imagehome.html
$Id: imagehome.html,v 1.9 2009/09/11 20:38:28 fn Exp $