Preprocessing buttons are arranged along the bottom of the image window:
Clicking each button will reveal a side panel on the image window with preprocessing options. Tools include
- smoothing using an isotropic Gaussian kernel (3D, 2D)
- ASL (arterial spin labeling) for smoothing, subtraction, averaging, and CBF quantification
- cropping to reduce the matrix size and FOV
- reslicing to make the matrix of an image series match another series with consistent coordinates
- partial volume correction (PVC) using the geometric transfer matrix to produce an ROI analysis
- alignment to a template
- brain extraction to remove extra-cerebral tissue
- motion-correction (work in progress; currently disabled)
Common features of preprocessing modules are
- A “run” button replaces the current time series with preprocessed time series within the program memory. No files are saved by this process.
- A “save” button will output the time series to one or more files with the same name (defined in a text box in the preprocessing page) but in the different directories that hold the time-series files. For instance, if one starts with a series of directories named “scan1”, “scan2”, …, and loads all the time-series files named “raw.nii” within those directories, then the output of smoothing will be “scan1/smooth.nii”, “scan2/smooth.nii”, … if the file name in the proprocessing page is “smooth”. Note that file names are “remembered” by the program for continuity across invocations of the executable.
Smoothing & cropping are self-explanatory. Smoothing via the k-space approach minimizes wrap-around by padding the data by one FWHM of the Gaussian smoothing kernel in the spatial domain using a reflection of the data about the edge of the FOV.
Reslicing refers to the process of rearranging voxels within the same coordinate space. In fastmap, the dataset loaded in the anatomy space forms the template, and the dataset loaded into the time-series space gets reformatted into the same voxel space as the template by assuming that coordinates are matched. Generally, it is not necessary to reslice one dataset into another for the purpose of alignment, unless one uses a different (non-fastmap) program for alignment; the fastmap alignment method includes a reslicing transformation as a default.