2 Data Conversion

When first working with an MR scanner, it may seem as if a MRI scanner directly acquires images of the object that lies inside its bore. Although images of this object can be seen on the scanner monitors, in its original form the raw MR scanner data is not a picture. Instead it resembles the frequency of the obtained signal and the location where the signal originated in the scanned object. The spatial-frequency space in which the data is sampled is referred to as k-space. The MRI physics video tutorials explain what k-space is. On most MRI scanners k-space data can be stored in a file format called DICOM, which stands for Digital Imaging and Communications in Medicine. Some manufacturers (e.g. Siemens, Philips, GE, Bruker) use their own file format (e.g. Philips uses PAR/REC format) to store k-space data.

Data conversion from k-space format to a real image is often a necessary data-preparation step and is almost always carried out before the data is being processed further. Some data-preprocessing can be done directly in k-space, but this is often not done. Most analysis software cannot directly read/process k-space data. Packages, such as Statistical Parametric Mapping (SPM), BrainVoyager, the FMRIB Software Library (FSL) and the Analysis of Functional NeuroImages (AFNI) all require data in real image space. The mathematical transformation that is used to change the frequency-space data into an image is called the Fourier transformation. Depending on the requirements of a particular software packages, the data may have to be converted into a different image space data, such as AFNI BRIK, ANALYZE, bshort, COR GIfTI, MINC(2) or NIfTI-1 format. Tutorial 2 in the fMRI section will explain how to convert data practically, pointing to some useful software applications and providing some data considerations.

Continue reading: Data preprocessing