Normalization question - ROI only or whole image - CT/CBCT

[daniloschneider]

Hello,
i am working on jaw bones CT images, from different scans and patients.
First, I thought to normalize the entire DICOM image, to reduce the variability interexams.
Second, extract the gray mean level and std dev of a delimited irregular ROI.
Third, extract texture features of that same ROI.
My question is:
1. Should i normalize (contrast enhancement) the entire image or only the ROI?
It seems that if I normalize the entire image is easier to delimitate the ROI (better contrast), but if I normalize only the ROI, the result (visually speaking, more gray range) for texture features extraction is better.
2. Can I make the normalization of the whole DICOM image and after normalize only the ROI?
in this case, we have two normalization steps in ROI.
I don´t if it is technically right and if i will have correct and reliable measures.
thanks,
Danilo Schneider

[Kenneth Sloan-2]

Excuse my fuzzy memory - but aren't CT images "born normalized"?  That is,
the gray values have semantic content.  Enhancing the contrast may make it easier to
see some things - but (as I recall) it loses information.
--
Kenneth Sloan
[hidden email]


--
Kenneth Sloan
[hidden email]


On Dec 8, 2013, at 08:24 , daniloschneider <[hidden email]>
wrote:

> Hello,
> i am working on jaw bones CT images, from different scans and patients.
> First, I thought to normalize the entire DICOM image, to reduce the
> variability interexams.
> Second, extract the gray mean level and std dev of a delimited irregular
> ROI.
> Third, extract texture features of that same ROI.
> My question is:
> 1. Should i normalize (contrast enhancement) the entire image or only the
> ROI?
> It seems that if I normalize the entire image is easier to delimitate the
> ROI (better contrast), but if I normalize only the ROI, the result (visually
> speaking, more gray range) for texture features extraction is better.
> 2. Can I make the normalization of the whole DICOM image and after normalize
> only the ROI?
> in this case, we have two normalization steps in ROI.
> I don´t if it is technically right and if i will have correct and reliable
> measures.
> thanks,
> Danilo Schneider
>
>
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Normalization-question-ROI-only-or-whole-image-CT-CBCT-tp5005843.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
Kenneth Sloan
[hidden email]

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Michael Doube-4]

On 09/12/13 03:12, Kenneth Sloan wrote:
> aren't CT images "born normalized"?

Most medical CT scanners output Hounsfield unit (HU) calibrated images. Sometimes the pixel value = HU, but more usually there's a linear calibration curve applied to raw pixel values. The main data points are air (-1000 HU) and water (0 HU). They're also usually at least 10-bit, often 12-bit or 16-bit data, so you can get a better visualisation by changing the brightness and contrast or window and level. There's no need to 'Apply' these to the data, just fiddle with the controls until you see what you want. Remember, most displays are 8-bit which matches well human dynamic response, but CT data have a greater bit depth which can be stretched to fit into that range.

If you are using X-ray microtomography to image your jaw bones, it's likely that your images aren't HU calibrated. However, the above still holds. Don't normalise, just adjust the window and level so you can see the contrast you need in the 16-bit data. But don't 'Apply'!

Michael

--
Dr Michael Doube
BPhil BVSc PhD MRCVS
Lecturer, Comparative Biomedical Sciences
The Royal Veterinary College, University of London
London NW1 0TU
United Kingdom

[phone] (Internal: 5503)

<http://www.rvc.ac.uk>

This message, together with any attachments, is intended for the stated addressee(s) only and may contain privileged or confidential information. Any views or opinions presented are solely those of the author and do not necessarily represent those of the Royal Veterinary College.

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[daniloschneider]

Thank you very much for the replies.