[MINC-users] problem with nu_estimate_np_and_em

[Andrew Janke]

> On Wed, Oct 1, 2008 at 5:41 AM, Tejaswini Ganapathi
> <tejasvinee at gmail.com> wrote:
>> All of my installed files (exe files) are in /usr/bin. Infact nu_correct,
>> nu_estimate and nu_evaluate work fine. I'm having this problem only with
>> nu_estimate_np_and_em
>>
>> Ive tried the following on the shell command line,
>> $ nu_estimate_np_and_em -normalize_field -save_fields -mask
>> mask.mnc breast.mnc fields_time.mnc
>>
>> nu_estimate_np_and_em: couldn't find program "class_statistics"
>> nu_estimate_np_and_em: couldn't find program "estimate"
>
> I reckon you have hit on a bug.  And amusingly I cannot find these two
> executables that you are talking about on my system (or yorick!)
> either. I suspect that you are using a little used part of the N3
> package that has escaped inspection/execution for a long time.

An update on this one, I have been in contact with the original Author
of N3 (John Sled) and he tells me that these errors are due to the
script in question trying to use another version of intensity
correction that is based upon the EM algorithm. nu_estimate_np_and_em
was written to compare the two methods.  He tells me that it is likely
not worth resurrecting the EM method (although I did find the code
hidden deep in an archive -- Thanks Mike and Sylvain). So the "fix"
for now is to not use the -mean or -tag options and be sure to use a
-sharpen x y option.

The default option for sharpen is:

   -sharpen 100 0.01

Mind you this he tells me is why nu_correct and nu_estimate were
written as they are simply wrappers for the bits of code in question
that make sure that only the NP method (or N3) is used.

There is now a "fix" in CVS for this that should make the next release
that will warn users of this.


-- 
Andrew Janke - andrew.janke at anu.edu.au
Department of Geriatric Medicine, ANU
(a.janke at gmail.com || http://a.janke.googlepages.com/)
Canberra->Australia    +61 (402) 700 883