Image normalization and standardization

You work on a certain topic or data set and don't know how to start off with it using MemBrain? Not sure if your net design matches your problem or if there is room for optimization? Is it reasonable at all to approach your problem with NNs? Is MemBrain the correct tool to accomplish your task and to match your infrastructure?

These questions are best placed here!
Post Reply
rgh
Posts: 2
Joined: Thu 3. May 2012, 21:02

Image normalization and standardization

Post by rgh »

Hi, I have a problem concerning MRI image normalization and scaling data between 0 and 1.
I have three tissue classes with different intensity range on a set of images (intensity range for each tissue is varying from image to image): for one image for instance I have: GM (100-150), L(0-150), WM(150-300). Now I want to normalize L to GM and WM, such that I can say how much brighter or darker L is relative to GM and WM. I was wondering how can I do it? and if I want to show the contrast of L, relative to GM and WM to be between 0 and 1, what will be the equation?
Thanks for your help in advance
User avatar
Admin
Site Admin
Posts: 438
Joined: Sun 16. Nov 2008, 18:21

Re: Image normalization and standardization

Post by Admin »

Hi,

I'm not quite sure if I fully understand the question: Do you know the intensity ranges for each picture and now want to tell how much darker or brighter L is?
In this case I suppose something like this is an approach:

ratio(l/GM) = (l - LL) * (LH - LL) / (GMH - GML)
ratio(l/WM) = (l - LL) * (LH - LL) / (WMH - WML)

where
l = Intensity of L in the region of interest
LH = High value of L range over whole image
LL = Low Value of L range over whole image
GMH = High value of GM range over whole image
GML = Low value of GM range over whole image
WMH = High value of GM range over whole image
WML = Low value of GM range over whole image

Just a guess, I'm not quite sure if I got your question correctly...
Thomas Jetter
rgh
Posts: 2
Joined: Thu 3. May 2012, 21:02

Re: Image normalization and standardization

Post by rgh »

My goal is to compare intensity of ROI "L" on a series of images with each other, but each image has a different intensity range.
So I need to normalize intensity of each image first to something on that image, that at the end I be able to show the contrast: so here I have WM, and GM ROIs.

Now to show how much intensity of L brighter or darker is than WM and GM (on the same graph) for each image. So calculating each ratio separately does not help me.

I was thinking of making WM and GM 0 and 1 respectively, so I can show the status of L between this two classes. And I was thinking two point normalization may help me and an equation like this might be helpful,
New L = (Median intensity of "L" ROI - Median intensity of GM ROI) * (1/ (Median intensity of WM ROI -Median intensity of GM ROI))

But my scale didn't become 0 and 1, bust something between -1 and 1 almost.
I don't know which part I did wrong
Post Reply