ROI Toolbox Docs

Documentation: Slice Repair (spike_repair.m)
--------------------------------------------
Paul Mazaika
12/10/04

1. Summary

   Reads a set of images and writes a new set of images 
   after filtering the data for noise.  User is asked to choose the
   repair method or methods. This program is best applied
   to the raw images, so the cleaned output images can be fed into
   slicetiming or realignment. When bad slice repair is chosen,
   a textfile BadSliceLog lists the repairs made.

   No original data is removed- the old images are all preserved.

2.Usage

 User is asked via GUI to supply:
   - Set of images
   - A repair method
   - If repairing bad slices, user can choose a threshold.

 POSSIBLE REPAIRS
 1. Filter all the images with a 3-point median filter in time. An 
     excellent filter for TR=2 or less.  If TR > 2, this filter
     may clobber activations in event-related experiments.
     Adds a prefix "f" to the cleaned output images.
 2. Detect and repair bad slices. Derives new values for the bad slice
     using linear interpolation of the before and after volumes.
     Bad slices are detected when the amount of data scattered outside
     the head is at least T above the usual amount for the slice. (The usual
     amount is determined as the average of the best two of the first three 
     volumes. It differs for each slice.) The default
     value of T is 5, which the user can adjust. A preview estimate
     of the amount of repaired data at T=5 is shown in the Matlab window.
     This filter removes outliers, but may reduce activations, so it's 
     safest when fewer than 5% of the slices are cleaned up. 
     Adds a prefix "g" to the cleaned output images.
 3. Eliminate data outside the head. Generates a head mask automatically,
     and writes it as ArtifactMask.img. Sets voxels outside the head 
     to zero. This process may help realignment to be more accurate.
     Adds a prefix "h" to the cleaned output images.
 4. Combinations of methods 1 and 3, or 2 and 3.
