sorry, bet'ed is not the right word - what I mean is that even when the
underlying image and the stats images have 0 value for all 'background'
voxels (i.e. have been betted) the output from overlay will have non-0
values in the background. This is due to the way that stats overlays
are represented within the block of data - basically overlays have a
single colormap (dk blue -> light blue, black->white, red->yellow) and
the three input images are scaled and shifted in value to the
corresponding ranges. fslview and slicer know what to do from the
aux_file field in the analyze header. You can check this easily by
loading the different images into fslview. 'avwstats file -w'
calculates the smallest ROI containing non-0 voxels, i.e. always give
you the full x-,y- and z- range on overlays as the background is no
longer at 0.
You need to calculate the correct range from the mask image in the
blah.ica folder, from the bet'ed underlying image or the thresh_stats
image. Once you have these numbers, you can crop the overlay images
(you don't have to crop the input volumes, you only need to get the
correct range from the inputs!)
Getting the colorbar is a bit more complicated - you can only get it
the way it looks in the ica report pages from melodic - this is because
when using it from the command line, slicer only sees the ovl image and
the information wrt the original intensities has been lost and we have
not yet written slicer code for supplying these numbers manually from
the command line.
You can, however, get a colorbar png from overlay itself:
overlay 1 0 bg -a map1 min1 max1 map2 min2 max2 cbar.png ysb ovl
will generate the ovl analyze file (in 'ovl') but also create a
colorbar PNG file (cbar.png) which is the 'yellow' range stacked on top
of the 'blue' range ('ysb') (if you don't want it stacked you can use
'yb' instead). You can 'attach' this to the slicer output with
pngappend sliceroutput.png + cbar.png combined_output.png
There is, however, nothing wrong with cropping the original data before
melodic - this should not change any of the results (the background 0
voxels are ignored within melodic anyways).
This will also sort out the colorbar problem and I really recommend
that you do it that way ... much easier.
Alternatively just use Opera and resize the entire web-pages (sometimes
a bit slow) or open all the png's in gimp (this will automatically
re-size) or (if on a mac) just use 'open
./blah.ica/report/IC_*thresh.png' and you can very quickly flick
through the spatial maps - this is what I do.
On 17 Dec 2004, at 03:46, Bettyann Chodkowski wrote:
>> (the reason why
>> it doesn't work on the overlay is that the background image is not
>> beted - if you get the parameters from the mask image it should work!)
> the background image in the overlay is, in fact, bet'd:
> the background image is one timepoint from filtered_func_data, the
> results of FEAT.
>> (iii) use avwroi to crop the original data before feeding into melodic
> i do not wish to change the results from melodic, per se. i only wish
> to change the .png images. i am hesitant to crop the input data to
> melodic. it seemed straightforward to crop the overlaid volume and
> then slicer that. but the bounding box returned by:
> avwstats <overlaid> -w
> returns the full extent of the volume, ie, no cropping. why is this?
> another approach is to crop the input volumes to overlay. this is a
> bit more cumbersome than simply cropping the resulting volume
> from overlay. does the overlay put a pixel in the corner of the volume
> as a way to deal with the colortable?
> another question that i cannot resolve: how do i include the
> colorbar in the .png? i hope this is not a secret.... :)
> thank you,
> - bettyann
Christian F. Beckmann
Oxford University Centre for Functional
Magnetic Resonance Imaging of the Brain,
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
Email: [log in to unmask] - http://www.fmrib.ox.ac.uk/~beckmann/
Phone: +44(0)1865 222782 Fax: +44(0)1865 222717