KCG Lab

F-MRI images are 4-dimensional images which depicts a time varying 3d clip. Such clips are then analyzed using dedicated software. These provide the computational capability required to reconstruct the large numbers of images and provide the statistical analysis tools to identify the anatomical regions that are active during specific tasks. As well as able to convert image formats that are used for MRI, FMRI, CT, PET scans (such as DICOM, TIFF, BMP, JPEG, Interfile, PNG, PGM, GIF, Raw Image Data, and other uncompressed image formats) into STL, (or AutoCAD DXF, 3D Studio (3DS), IGES, VRML, Wavefront OBJ, raw triangles, and other 3D graphics file formats for rapid prototyping, 3D printing, animation and visualization applications). Still those clips require >40 MB each which make analysis time consuming. Furthermore image processing procedures based on compressed clip slices (such as JPEG) suffer from data loss.

The research project makes use of Nvidia's GPGPU platform (CUDA) in order to provide an efficient algorithm for 3D-4D-image compression. Our suggested compression scheme consists on

• Computing (parallel wise) the discrete cosine transform of 4D blocks.
• Filtering out (parallel wise) non-used frequencies. i.e. those that physicians and brain scientist find not important.
• Reconstruction of the fragmented image.