[MINC-users] Problem with mritotal

[Claude LEPAGE]

Hi,

Some notes about the history of bestlinreg.pl:

- I created this script as a simplified version of mritotal for linear registration
  using a brain mask, in the same way that nlpfit was created for non-linear
  registration -- because it was too difficult to modify the do-it-all script
  mritotal. Remember that mritotal can do both linear and non-linear registration. 
- bestlinreg.pl basically reproduces the same fitting steps as mritotal - including
  blurring as in mritotal.
- unlike mritotal, bestlinreg.pl does not have a config file (good or bad???)
- bestlinreg.pl can be used without masking, if desired
- Following Louis' suggestion, I tried linear registration without blurring (and
  also without the gradient image), but the results were not convincing - too many
  failures.

>From the thousands of scans processed with bestlinreg.pl, I can report very few
failures. This was not the case for mritotal where often the subject's head ended
up being fully inside the skull of the model (hence the smaller head).

To the question if mritotal should no longer be used, my answer is if the code
is too "monstruous" to be maintainable, then the answer is use simpler scripts that
perform the same tasks but that can be modified and maintained easily. Modularity,
simple tasks... the key to structured programming. (Sorry if I offended anyone.)

As for releasing bestlinreg.pl, I'll leave it in Andrew's hands to decide if it
should be part of the next release of mni_autoreg. I think it should.

Claude