Dear all,

What tools do people recommend for visualising

(a) overlays or blends of two (or a few) 3D volumes, e.g. two opposite thresholded t-maps on a group average T1; or overlaying a non-rigid reg Jacobian determinant image (not thresholded) on its corresponding target image.

(b) 4D time-series of more than a few 3D images

(c) 4D (or 5D with trivial 4th dim, in NIfTI) vector fields, e.g. transformation or displacement fields, visualised as arrows, and/or with various colour schemes

(d) ditto for tensor fields, e.g. DTI ellipses


To get the ball rolling, my current thoughts:

(a) FSLview can do thresholded overlays, and can vary the transparency of these, I personally don't know how this compares with other truecolour blending schemes in terms of how multiple semi-transparent colour overlays combine; I don't think colour scale-bars can be automatically added. SPM5's orthviews ("check reg") and slover (http://imaging.mrc-cbu.cam.ac.uk/imaging...)
are nice for overlays or blends, and add colour scale-bars automatically. SPM however doesn't have FSLview's snappy interaction (click new cross-hair position, wait for jump, rather than drag cross-hairs around in real-time) and of course, requires a MATLAB licence.

(b) FSLview has a nice movie mode (if you have enough RAM for your complete 4D data), and the ability to graph the current voxel's value over the time-series; its lightbox mode is only for multiple spatial slices of a single volume (though you can "movie" them through the time-series) and not multiple volumes at a single spatial slice. SPM's orthviews in contrast provides the latter but not the former; it's also limited to viewing <=15 images at a time. Christian Gaser has an extra for SPM (SPM2 only?) which can show many more slices (one fixed plane only, not three tied ortho views, and not easily changeable): http://dbm.neuro.uni-jena.de/vbm/display...

(c) I don't think either FSL or SPM's viewer support 5D NIfTIs for vector fields as the standard suggests (page 4 of http://nifti.nimh.nih.gov/nifti-1/docume... has a nice summary of this). FSLview can display vector fields stored in 4D NIfTIs (dim4=3) as lines or direction-colour-coded overlays (http://www.fmrib.ox.ac.uk/fsl/fslview/dt...). Because it's meant for DTI rather than displacement fields, it doesn't represent the sign/sense of the vectors (e.g. with arrow-heads). I have no experience with DTI, but I don't think FSLview renders tensor ellipses. SPM can use three separate 3D images or a 4D image (as FSL) for a vector field, offering an RGB overlay. SPM seems very slow doing this though. There are "quiver" and "quiver3d" orthview plugins, which presumably draw lines like FSLview, but they don't seem to work for me (on a 4D displacement field, which works with RGB overlay). John Ashburner posted some MATLAB code to do this to the SPM mailing list (http://www.jiscmail.ac.uk/cgi-bin/webadm...), but it's a long way from the "simplest" way this forum is looking for. I don't think either FSL or SPM can colour-code 2D (in-plane) lines based on the third (out-of-plane) component, as I believe is sometimes done in DTI/tractography. What are people's thoughts on this or other visual representations -- with regard to simple tools which offer such options?

(d) As (c), 5D NIfTIs are not supported. I don't think either FSL or SPM can illustrate tensor fields in any direct way (a derived principle eigenvector can be viewed as above, but not directly from a tensor data-set). I don't think tensor ellipses can be illustrated in either program, but possibly I am missing something...

Mango (http://ric.uthscsa.edu/mango/) was announced recently in this forum, it looks like it should be very nice for (a), and probably also for (b) though I'm not sure how many images it can stack, but I can't see from its website if it offers anything for (c) or (d) -- Angie?

Well, that's more than enough from me. I look forward to hearing other people's thoughts,

Ged