Hi Emperor Zhark,
Thank you again for implementing the statistic-as-alpha feature (and so quickly!). It has already become my default view for browsing functional datasets, and I was thinking, in the immortal words of Tom Lehrer, "More, more, I'm still not satisfied!" I mean... um, it would be
really awesome if I could use the Clusterize plugin with AFNI_FUNC_ALPHA turned on.
The main way this would help is turning this view:
p < .005, alpha < .05
"Hey look! We found something! Let's write a paper!"
...into this view:
Full opacity at p < .005, white outline around clusters with alpha < .05, black outline around
a priori mask
"Huh. Maybe the effect we're seeing here is a different effect from the one we were expecting to see. Let's write a paper!"
Producing this second graphic involved saving several different PPMs with the AFNI viewer, then munging them together with numpy and finally converting to PNG. I tried to automate the AFNI viewer part using plugout_drive, but it turns out that the Clusterize plugin doesn't register any driver functions, so I'd have to use 3dClust offline, and in the end the easiest way was to use the viewer interactively. And while
I am willing to spend way too much time working out the details of the PPM file format and how they interact with numpy.fromstring, most AFNI users are not quite that crazy.
You mentioned before that this would not be as easy a modification to make as AFNI_FUNC_ALPHA, but I think it would be worth it. The difficulty in specifying the interaction seems to be resolved by taking one outline around the CLUST_MASK and one around the regions that Clusterize would currently display, though as you can see in the figure, this doesn't quite interact well with resampling the overlay to match the underlay resolution. I have several ideas on how to handle that and none of them are any good.
ijs
Edited 1 time(s). Last edit at 02/20/2015 12:51PM by Isaac Schwabacher.