Rescale Intercept and Slope tags...
Gonzalo Rojas
I transferred some MR images to a CD-R in DICOM format (using EasyVision
workstation)... That images were acquired with a Philips Gyroscan Intera
MR... I need those images to do some quantification using image data but I
have a doubt (and for that reason, I need exactly the image data acquired
in the MR system): it seems to me (visually) that the images were saved
with a lossy method... I looked the DICOM tags in more detail, and I found
that "Rescale Intercept" (28/1952) and "Rescale Slope" (28/1053) tags have
the values (respectively): 0.000000 and 22.619610... What it means both
tags?... If anyone is interested to look in more detail, I could send one
of the DICOM files...
Thanks in advance...
Gonzalo Rojas
Somchai
> that "Rescale Intercept" (28/1952) and "Rescale Slope" (28/1053) tags have
> the values (respectively): 0.000000 and 22.619610... What it means both
> tags?... If anyone is interested to look in more detail, I could send one
> of the DICOM files...
The value of pixel is the result of data in (7FE0,0010) multiply with
"Rescale Slope" (28/1053) plus "Rescale Intercept" (28/1952).
Pixel(x,y) = Data(x, y) * Rescale Slope + Rescale Intercept
In MR, "Rescale Slope" can vary from slice to slice because
MR doesn't have standard scaling like Housfield Unit in CT.
Dan Konigsbach
processing for some types of images. Rescale operates on stored pixel
values, which are the pixel numbers that you get directly from an
uncompressed DICOM image, or that you get out of the decompressor for
a compressed DICOM image.when decompressing the image).
Rescale is designed for modalities where the pixel values are
meaningful measurements, not just arbitrary numbers, e.g.:
Hounsfield units for CT,
Optical Density of CR,
Becquerels/milliliter, counts, or one of many other types of
values for PET
(the list could go on)
These measurement units are not necessarily the most convenient or
efficient for a modality to work with or store in the DICOM image.
Instead, the modality uses the Rescale values to show how to convert
from stored pixel values to meaningful units. Rescale is a linear
transformation:
If:
x is the stored pixel value
m is the value of Rescale Slope (0028,1053)
b is the value of Rescale Intercept (0028,1052)
y is the rescaled, meaningful value
then
y = (m * x) + b
If there is a rescale, then Window Width and Window Level are
specified in rescaled units. This means that you first apply rescale
to the pixel value, then window/level.
The MR SOP class does not allow images to have a Rescale Slope or
Rescale Intercept, since MR pixels values don't correspond to some
specific physical measurement. But...
When you lossy-compress an image, you often need to scale the pixels.
The official "right" way to do handle this is to make a compensating
adjustment to the Window Width and Window Level, but some prefer to
keep the Window Width and Level the same in the compressed and
uncompressed image. Instead, we put the the adjustment in the
Rescale. Legal? No. Done? Yes.
Some more notes:
* Rescale is a linear operation. If that's doesn't do the job, an
image can contain a pixel lookup table, called a Modality LUT,
instead. An image can have a rescale, a modality LUT, or neither. It
can't have both. The DICOM module that holds either the Rescale or
Modality LUT is itself called the Modlaity LUT Module.
* For the X-Ray IODs, the rules get more complicated. For them,
trust nothing in this message.
* You think these image have lossy compression. Here's how to be
sure:
- The Transfer Syntax will tell you if the image is currently
compressed.
- Lossy Image Compression (0028,2110) will tell you if the image has
ever been lossy-compressed.
Sorry this is so long and complicated. I hope it helps.
- Dan
Gonzalo Rojas <rojas_gon...@hotmail.com> wrote in message news:<3e419816$1...@nova.entelchile.net>...
jose manjon
jose
Dan Konigsbach wrote:
--
##########################################################
Prof. Jose Vicente Manjón Herrera
Dept. Fisica Aplicada
Escuela Universitaria de Informatica
Universidad Politécnica de Valencia (SPAIN)
##########################################################
Gonzalo Rojas
news:YJJ0a.1834$Rq4....@newssvr16.news.prodigy.com:
Hi:
But what is the meaning (or reason to use) the rescale slope tag ?... In
other words, what is the difference between a DICOM image saved with a
Rescale Slope equal to 1.0 (or without that tag), and a DICOM file with a
Rescale Slope tag greater than 1.0 ?... And, how can I get the "original"
(not processed) RAW image data from such a DICOM file ?...
Gonzalo Rojas
news:3E43A096...@fis.upv.es:
Hi:
Reading Dan's e-mail, I understand that Rescale slope tag it is used in
MR images to create a pseudo-compressed image only...
Best regards,
Gonzalo Rojas
Dan Konigsbach
creating a pseudo-compressed image. Rather, I meant to say that some
folks add a rescale to MR images when doing (actual) compressing so
that the compressed image uses the same window width and level as the
uncompressed image.
So, must you use rescale in MR?
If you're creating an uncompressed MR image, I suggest that you play
by the rules and don't use rescale.
If you're compressing an MR image, and if you need to scale the pixel
values to get decent compression, it's between you and your conscience
whether to add rescale or to modify the window width and level.
If you're trying to view an MR image and it has both rescale and
window values (or a display LUT), then the window (or LUT) values are
based on the rescale - you have to apply the rescale before you apply
the window. (Alternatively, you can just say that the message is
illegal and complain, but I'd suggest leniency here.)
I hope this helps.
- Dan
Gonzalo Rojas <rojas_gon...@hotmail.com> wrote in message news:<3e43e8d4$1...@nova.entelchile.net>...
> jose manjon <jma...@fis.upv.es> wrote in
> news:3E43A096...@fis.upv.es:
>
> > Finally, for MR images must we use Rescale?
> >
> > jose
> >
> >
> > <snip/>
Gonzalo Rojas
news:9b9c7463.03021...@posting.google.com:
Hi:
What it means the tags window width and level ?... What are the use of
them ?...
Best regards,
Gonzalo Rojas
jose manjon
(convert your DICOM data to gray bitmap (0-255)).
if you have a window center 1000 and window width 500 it means that
the pairs (1000-500,1000+500) corresponds to (0-255) bitmap values.
jose
Gonzalo Rojas wrote:
--
Gonzalo Rojas
> is the values proposed to get the same visualization on every place
> (convert your DICOM data to gray bitmap (0-255)).
>
> if you have a window center 1000 and window width 500 it means that
> the pairs (1000-500,1000+500) corresponds to (0-255) bitmap values.
>
> jose
Hi:
In your example: if I have a "window center" of 1000 and a "window
width" of 500 (1000-500,1000+500), and "Bits stored" is 8, then I don't
have all the information of the image... I lost some image information when
I saved it... To have all the image information, I need a DICOM file of at
least 12 bits/pixel... Is it true ?...
Best regards,
Gonzalo Rojas
jose manjon
anyway you need no to include all data for getting a valuable image.
I mean, if you have 8 bit stored image you have a range (0,255) and the whole
range window is (WC=128,WW=128) but if its proposed other window is because in
this way you can get the same visualization than in adquisition time.
if you get a 12 bit image, your range is 0-4095 and the same thing can be done.
modify window paramenter play a important role on image visualization and this
is a common operation on all medical image software.
jose
Gonzalo Rojas wrote:
--
Gonzalo Rojas
Hi Jose:
But I need all the "numerical" information in the image, because we need
to do some quantification using the data in the images... For that reason I
am doing all this questions...
Best regards,
Gonzalo Rojas