Prev: GPS simulink model
Next: 1-D advection-diffusion: how to benchmark.
From: amit on 21 Mar 2010 04:18
i m a final year student i m doin a project"texture analysis of biometric images"
my project is abt image processing .i m using MRI image of brain as input after processing it(i.e.removal of noise,filtration) i m using fuzzy clustering algo to locate the exact location of tumor in the brain.plz help me
reply asap
From: Walter Roberson on 21 Mar 2010 04:48
amit wrote:
> i m a final year student i m doin a project"texture analysis of
> biometric images"
> my project is abt image processing .i m using MRI image of brain as
> input after processing it(i.e.removal of noise,filtration) i m using
> fuzzy clustering algo to locate the exact location of tumor in the
> brain.

That approach will have low sensitivity, being able to detect tumors
only after they have grown enough to distort the typical brain structure
patterns beyond the natural deviation between people. For high
sensitivity, MRS is *much* better than MRI in locating tumors before
they are even visible structurally. As the tumor progresses but before
it has grown too large, you will probably get color changes (due to
chemical changes) before you get physical distortions, so _texture_ is
not the first choice of image features to investigate.

Are the images 2D MRI or 3D MRI? Typical clinical 3D MRI is done at too
low of a resolution to have anything but the crudest features -- and the
pulse sequence for 3D MRI usually isn't tuned to be an especially good
detector of tumors. 2D MRI that is designed to find tumors is usually
done as a time course (sequence of images of the same area), and the
first thing that is usually searched for in the images is areas of
unusually high blood flow and unusually high concentration of blood
vessels -- because the tumor is hungry to be fed. Blood flow can be
approximated by oxygen flow, but there is a way to tune for oxygen and
iron together that does much better.

I don't know if we've ever done texture analysis, but we've done a lot
of brain MRI work. I'd like to mention some of the things we're working
on now, but I'm not authorized to talk about them :(

Two key places to look for relevant information would be the Journal of
Brain Medicine (JBM) -- but that's very expensive, I understand -- and
the Proceedings of ISMRM.
From: Rune Allnor on 21 Mar 2010 07:29
On 21 Mar, 09:18, "amit " <amit_mishra...(a)yahoo.com> wrote:
> i m a final year student i m doin a project"texture analysis of biometric images"
> my project is abt image processing .i m using MRI image of brain as input after processing it(i.e.removal of noise,filtration) i m using fuzzy clustering algo to locate the exact location of tumor in the brain.plz  help me
> reply asap

In other words, the goal of your project is to

1) Detect the presence of a feature in the brain
2) Classify the feature unambiguously as a tumor
3) Pin-point the position of this tumor

First of all, make sure you have some serious insurances
against claims of malpractice. Second, make sure you post
who your clients / employers are - I for one would certainly
*not* have surgery done by somebody who relies on codes
obtained from USENET fora...

Rune
 | 
Pages: 1
Prev: GPS simulink model
Next: 1-D advection-diffusion: how to benchmark.