Preprocessing

sMRI Preprocessing

Since sMRI stores information about tissue types, it is necessary to segment structural images into separate different structural tissues such as gray matter, white matter, cerebrospinal fluid (CSF), blood, bones, etc. The segmented images are used later for fMRI coregistration to a common space

fMRI Preprocessing

Functional images can be acquired in two frequency bands: single-band, or multi-band. Single band acquisition uses longer repetition time (TR) and thus takes longer duration of acquisition. However, older MRI scanners may not have access to multi-band sequencing. For single-band fMRI, it is necessary to correct motion artifacts in time according to a reference slice. Then, the images are realigned to correct motion artifacts in space. The functional images are then coregistered with sMRI and normalized to the Montreal Neurological Institute (MNI) Space for better mapping of brain regions. Smoothing of functional images increases signal-to-noise ratio (SNR) by taking the weighted average of the neighboring pixels

dMRI Preprocessing

Diffusion-based images first undergo eddy current correction. Eddy currents are void-like structures in the images that are induced by the external magnetic field during acquisition. These are removed to improve accuracy of the tract-based analysis. Then, the first slice is extracted assuming that this slice is acquired when the magnetic field strength is still zero, The extracted slice is used as a mask for skullstripping, to remove the bones and facial structures. Finally, a tensor model is fitted utilizing the given eigenvectors of directionality

PET/CT Preprocessing

PET images are 3D objects that store information about metabolic activity of different regions of the brain. PET images can be treated similar to functional images except that the preprocessing does not involve slice timing correction. Below are the steps in preprocessing PET images