NIS (NeuroImaging Software) package version 3.1   September 30, 1999


Changes from NIS 3.0

- Bug fix to nisnrm (line 130, p++ should not be there but further down).
  Also fixed default target image intensity to be 3000, not zero.
- Fixed do_smooth so that it reads from the func_norm directory (it was
  previously reading from the func_reslice directory).
- Renamed parameter files for group analysis scripts from
  shrink_reslice_params and shrink_orig_params to do_params1 and
  do_params2 (so that a cp of do_* will grab everything needed).
- Changed do_ scripts to use absolute pathnames for subject and structs
  directories and to use xargs and recursive gzip/gunzip for greater
  reliability (greatly minimizing "arg list too long" errors).
- Fixed numeric overflow problem with image thresholds in niscrit.
- Fixed long-standing -T/-t bug in nispost4; removed nispost3 and
  renamed nispost4 to nispost.  Removed -a flag (effectively setting
  alpha to 1), and changed program to set below-threshold pixels to zero
  in output image and to allow negative t values.
- Added -f flag (input file) to nisanova, niscrit, nisnorm, and nispost.
- Added -d flag (output directory) to nisanova, nisnorm, nisoverlay,
  nispost, nisroi, and nistime.
- Added -o flag (output file) to niscrit; removed -o flag from nisoverlay
  (to eliminate potential conflicts with -d flag).
- Increased MAXFILES limit from 20000 to 40000.
- Added NIS java applications to package.
- Fixed read_analyze to read gzipped (.gz) as well as compressed (.Z)
  and zipped (.z) files.  Previously, it could only read .Z files.
- Changed iscompressed function to no longer assume .img suffix,
  which was causing a problem when iscompressed was called on a header
  file and the compressed status of the header and the image files
  differed (i.e. one was compressed and the other was not).
- Fixed problem with read_analyze: non-existent images were being treated
  as though they were compressed, resulting in mysterious failures; now,
  a "file does not exist" error is given instead.  Also, the compression
  status of the image is no longer assumed to be the same as the header's.
- Changed nisanova, niscrit, nisnorm, nispost, and unixstat_anova so that
  space for image filenames is dynamically allocated instead of defining
  a large (~40MB) two-dimensional char array for these names.
- Removed the use of a temporary file (anovaXXXXX) by unixstat_anova; it
  now uses an array in memory to hold factors and levels, which improves
  performance.  API is same as before.
- In nisanova, moved the computation of pixel sums (to check thresholds)
  outside of the pixel loop, to improve performance.


Changes to nisair:

- Added -a flag to specify output directory for airfiles (default is
  air_files).
- Added -l flag to specify output directory for log files (default is
  current working directory; default used to be the air_files directory).
- Added -r flag to indicate reslice is desired and to specify directory
  for reslice files (formerly defaulted to air_images.3D and performed
  reslicing automatically).
- Added -k flag to disable interpolation to cubic voxels on reslice
  (formerly this was hardcoded).
- Added -T flag to set temporary directory (formerly hardcoded as tempdir).
- changed -b flag to only specify FWHM for x and y dimension smoothing
  (previously the -b flag specified FWHM for x, y, and z dimensions).
- Added -B flag to specify separate FWHM for z dimension smoothing.
- Added -c flag for convergence threshold for iteration to cease (default
  is 0.0001).
- Added -s flag for initial sampling factor for pixels (default is 27).

NOTE: you must add the following flags to obtain the previous behavior:
     -k -r air_images.3D (if 3-D) -B <whatever was used for -b flag>
     -l air_files -T tempdir