ICs identification analysis

[Giacomo Tarroni]

Hello,

I've been recently using MIA to perform the non-linear registration of myocardial perfusion MR images, specifically by means of the algorithm mia-2dmyoica-nonrigid-parallel. I'd like some clarifications about the outcome of the --save-feature option:

1) How can I tell which component was identified by the ICA analysis as LV cavity, RV cavity, perfusion and baseline, respectively?
2) Shall I pay attention to the png and txt files tagged as "final" or to the other ones?
3) If the ICA was successful, the coefficients reported in the respective txt should depict the intensity values of the different ICs throughout the sequence, right?

Thanks in advance!

Giacomo

[Gert Wollny]

1. When you run the program with -V message the mapping will be printed out in each registration pass. Apart from that, if you enable cropping (i.e. --segscale > 0), then files <feature>RVic.png, <feature>LVic.png, <feature>RV_candidates.png, <feature>LV_candidates.png, etc will be created, that show what the feature images and the actual segmentation looks like. If you just want to see what the ICA segmentation and the reference images look like, then you might run mia-2dmyocard-icaseries, it calls the ICA without running the registration.

2. The final ones are evaluated after registration, hence you should first pay attention to the other ones.

3. The coefficients correspond to the ICA mixing matrix, i.e. the proportion by which each of the feature images contributes to the image at the given time step when reconstructing the series.

Hope that helps.
Gert