----- Original Message -----
From: "Rainer Heintzmann" <rheintz@gwdg.de>
To: "Roy Schestowitz" <roy@schestowitz.com>
Sent: Tuesday, December 07, 2004 1:49 PM
Subject: Re: AART
Dear Roy,
yes, I put in the username ftp@schestowitz.com and no password.
Logging into your system works fine, and I can change to the
public/AART2.0.0/ directory,
but since I do not know the filename, I cannot download anything.
I have checked the configurations again. I can't see what is causing this.
"dir" or "ls" in ftp does not seem to respond and I get a timeout after
300 sec.
I can actually see the html code, but I would have to download many files
individually...
Yes. That's why I set up the FTP 2 weeks ago.
Can you tell me, where to find the VXL registration-modules you described?
I'm afraid that this module is Manchester-only. You can find VXL at:
http://vxl.sf.net/
In my data I have something like 400 landmarks (from automatic image
segmentation)
which have to be matched to another in a time-series (80 timepoints, each
with ~400 markers) in an error-tolerant way.
Sounds like an easy task, but I am not sure how to proceed.
I think you might find SPM (from UCL) useful and I know the code is publicly
available as well. What you describe sounds accomplishable.
Error tolerant means, that a few new
markers are generated (new cells are "born") and old ones can dye.
That's where statistical models become problematic. You might have to add
and remove markers in an ad-hoc way.
Any ideas how to compute the best match? Sounds like graph theory to me...
Doesn't quite sound like graph theory to me... are you trying to find best
one-to-one mapping of markers/points?
Greetings,
Rainer
|
|
