Click on figure.
The process of volume rendering has three major steps. The process begins by preprocessing the input data, classifying it into values that are significant to the rendering process and then projecting each of the voxels into the final image.
The MRI data was collected using a Philips 1.5T Gyroscan. This is simply 12 bit voxel data. During the classification step of the rendering process, this data is colored using lookup classification tables. Essentially, the different ranges of input values are mapped to the different colors with the lowest values mapped to black, then red, followed by yellow and highest values are mapped to white. (This is commonly known as the "hot metal" color scale.)
The data was preprocessed to enhance the existence of surfaces in the data as well as to remove a rectangular region from the data volume. This allows us to 'cut' into the volume, permitting us to view any structure of interest.
Click on figure.
This film loop shows the capabilities of the PIXAR ChapVolumes subroutine library. The input data consists of 64 slices of 256x256 MR tomographic images. Preprocessing of the image included interpolation of the data up to 128 slices, surface extraction, region removal and pseudo coloring the data. The coloring was achieved using the VIPMAP subroutine with a 'hot metal' color table.
Each frame of the film loop was then created by projecting the processed data slice by slice, changing the rotational view angle for each frame by only a small amount.
A large amount of detail can be seen in this image that would be otherwise very difficult to view in the traditional slice by slice analysis of tomographic data. The spatial relation of the different structures becomes evident when view with this technique.
Click on figure.
This atlas is constructed by identifying 120 structures on a normal brain. This film represents a surface rendering of the atlas data where the individual structures have been clustered together to form the major lobes of the brain. Red denotes the frontal lobe, green denotes the parietal lobe, blue denotes the temporal lobe, and yellow denotes the occipital lobe.
Click on figure.
The PET data used as input for this image consisted of 18 adjacent planes spaces at 6mm intervals with a slice thickness of 12mm and a transverse resolution of 11-12 mm. This data was then interpolated along the three axis in order to have a 256x256x128 volume of data. Each voxel is represented by a 8 bit number.
The volume was then surface rendered using technique similar to those used in the MR rendering. PET data is usually very smooth, and surfaces are not very well defined. Because of this, the exterior surface of the PET volume appears to haves holes in it.
Click on figure.
The data was collected using the Inflow technique with pre-saturation pulses to image arterial and venous flow separately on a Philips Gyroscan 1.5 T MR system. The data is courtesy of Philips Medical Systems of Holland.
The two data volumes were then merged for the rendering process. Before merging, the two volumes were colored separately to show upflowing blood (arteries) as red and downflowing blood (veins) as blue.
Click on figure.
The PET data used as input for this image consisted of 18 adjacent planes spaces at 6mm intervals with a slice thickness of 12mm and a transverse resolution of 11-12 mm. This data was then interpolated along the three axis in order to have a 256x256x128 volume of data. Each voxel is represented by a 8 bit number.
This data set was registered to the multi slice MRI image of the same patient and interpolated to the same number of slices in the MR data Each slice consists of the PET activity shown in red overlayed on the anatomy shown in black and white by the MRI slices.
Click on
figure.
This is a 3-D volume rendering of an inflow MRA data set. The individual slices demonstrate both vessels and anatomy. The anatomy was rendered using the volume rendering technique described above and combined with the vascular information which was extracted from the original images and processed separately. The resulting image demonstrates both cortical vessels or both cortical and internal vessels.
Click on
figure.
This image is the same as above (section 7) with only cortical vessels shown.
Click on figure.
Combined 3-D MRI and PET activation data showing a deep region involved in pain processing. The MRI volume, shown in black-and-white, has been processed, or rendered, to enhance the 3-D information in the data.The PET data, shown as a red spot, were obtained using 0-15 labelled water to measure cerebral blood flow (CBF). A CBF image volume was measured in a baseline state, with a 42 degree thermistor placed on the forearm, and in an activation state, with the thermistor temperature raised to a somewhat painful (47/48 degrees) level. The CHANGE in CBF marks the brain region involved in processing the painful stimulus. Here, the same experiment was performed in 10 subjects and the results combined to increase their reliability. The region shown is in the anterior cingulate cortex (area 24) and has been exposed by removal of a sector of the 3-D MRI/PET volume. This region has been identified with pain in the past by surgeons who found that removing it had a beneficial effect on patients with intractable pain. There were in fact 3 other regions involved, in somatosensory cortex, but these are not shown because of the removal of the sector.
Click on figure.
This example of combined 3-D MRI/PET image data obtained during an activation study shows up parts of the visual cortex. Using the same CBF-subtraction methodology as before, the baseline condition had subjects view a simple white cross on a black background, the activation condition was to view black-and-white line drawings of animals. In this case the primary visual cortex (V1) was activated in both states but the activation condition recruited higher-order areas (V2) to process the increased feature content in the line drawings (i.e. edges,lines etc.). Thus the image shows the difference , i.e. increase, of CBF (in red) in this associative visual cortex at the occipital (rear) part of the brain.