[ Given the recent mail-server problems and that I've not seen my
message on the list - here it is again - apologies if it's a repeat ]

Dear Mary,

Mary Teruel, Dept. of Cell Biology DUMC wrote:
> Hi - Does anyone have recommendations for a good 3D imaging software package that can run on a Pentium based system? We would like to be able to manipulate stacks of 2D images (say 100-200 images at 1024 x 1024 resolution). We would like to use the software package to make some 3D measurements, but we are most interested in that it gives good 3D renderings of the data. As added features, it would nice to be able to do full rotations amd stereo pairs.

Lance Ladic maintains a nice list of 2D and 3D shareware for use with
confocal images.  See http://www.cs.ubc.ca/spider/ladic/software.html

Also the 3D reconstruction page has some useful pointers but it's
starting to get out-of-date.  See:

http://biocomp.arc.nasa.gov/3dreconstruction/

We use AVS/Express (http://www.avs.com), which I have found to be
excellent.  There is a steep learning curve (it's no turn-key system)
and requires some programming ability but is very extensible.

>      We have not purchased the Pentium system yet, and so we are open to getting a motherboard with dual or quad processors - or to getting accelerator cards.

The processor is important for sheer speed.  However, not all software
will be optimized for multi-(dual/quad)processors.  Once you've decided
on the software you want you will be able to determine whether
multi-processors will be of use.

More important is the issue of RAM:

> ... say 100-200 images at 1024 x 1024 resolution ...

i.e. 100-200 Mbyte images (more if they're greater than 8-bit).  So I'd
recommend you aim for at least a 256Mbyte system (we run 512Mbytes).
Failing that get a large disc with a _big_ swap partition so you at
least have alot of virtual memory.  This also assumes the software is
running in an operating system with good (virtual) memory management.  A
straw poll of local hackers ranks the various OSs thus:

Win3.x (forget it)
Win95  (fair)
WinNT / OS2 / Linux (good)

>      Does anyone have any comments on the Visilog program (from Noesis up in Canada)? It looks like a nice program, and it also can run with a Matrox accelerator card to really speed it up. However, they said their 3D software isn't "voxel" yet. Is this a problem?

I'm not familiar with Visilog but you will almost definitely need a
voxel-based rendering engine.

As for graphics (and other accelerator) cards, again this will depend on
your choice of software (see multi-processor comments above).

Good luck,
________________________________________________________________________
Chris Pudney              Biomedical Confocal Microscopy Research Centre
[log in to unmask]           Department of Pharmacology
http://www.cs.uwa.edu.au/~chrisp         University of Western Australia
PH (618) 9346 4571  FX (618) 9346 3469      Nedlands, WA 6907, AUSTRALIA