# Pixel Aspect Ration vs Pixel Spacing - DICOM

This is a discussion on Pixel Aspect Ration vs Pixel Spacing - DICOM ; Hi, everybody ... I read, at page 291, PS3.3-2006 (Information Objet Definitions) that, for US images e.g. V=0.30 mm, H=0.25mm Pixel Aspect Ratio = Vertical Size\Horizontal Size = 0.30mm \0.25 mm Thus the Pixel Aspect Ratio could be represented as ...

# Thread: Pixel Aspect Ration vs Pixel Spacing

1. ## Pixel Aspect Ration vs Pixel Spacing

Hi, everybody ...

I read, at page 291, PS3.3-2006 (Information Objet Definitions) that,
for US images
e.g.
V=0.30 mm, H=0.25mm
Pixel Aspect Ratio = Vertical Size\Horizontal Size = 0.30mm \0.25 mm
Thus the Pixel Aspect Ratio could be represented as the string "6\5",
or "60\50" or any equivalent integer ratio.

OK.

"12\10", should work, too, right?
How can I deduce that the Spacing is 0.3 * 0.25, and not 0.6*0.5?

Any information welcome.

Jean-Pierre Roux

2. ## Re: Pixel Aspect Ration vs Pixel Spacing

jpr@creatis.univ-lyon1.fr wrote:

> "12\10", should work, too, right?
> How can I deduce that the Spacing is 0.3 * 0.25, and not 0.6*0.5?

After browsing the archive, the only answer that actually make sense is
one made by D. Clunie:

....
I think it is also fairly unusual for viewers and displays
to handle non-square pixels "properly" anyway.
....

The best approach could be to simply set spacing of 1\1 and produce a
warning.

Mathieu

3. ## Re: Pixel Aspect Ration vs Pixel Spacing

Mathieu Malaterre wrote:
> jpr@creatis.univ-lyon1.fr wrote:
>
> > "12\10", should work, too, right?

Yes, of course.

> > How can I deduce that the Spacing is 0.3 * 0.25, and not 0.6*0.5?

You can't. The ratio is less information than the pair of pixel
spacings.

>
> After browsing the archive, the only answer that actually make sense is
> one made by D. Clunie:
>
> ...
> I think it is also fairly unusual for viewers and displays
> to handle non-square pixels "properly" anyway.
> ...

That's a bit pessimistic! You can easily get non-square pixels by
doing MPR; imagine an axial CT with 0.5 mm square pixels and 1mm slice
separation -- the orthogonal views will have 0.5 mm by 1 mm pixels.
Our viewer handles them fine. I'd be shocked if other commercial
viewers did not.

-Steve

4. ## Re: Pixel Aspect Ration vs Pixel Spacing

As a matter of fact, our customers have had numerous occasions where
non-square pixels were not being displayed correctly. Specifically,
the other vendors were only using Pixel Aspect Ratio, which works for
most of the IODs out there. But, CT/MR/PET... use Pixel Spacing, not
Pixel Aspect Ratio, which means that the images that were displayed
looked "squished"... Ultimately, we were forced to "fix" our images,
and make our pixels square, to satisfy the customer complaints.

Steve Robbins wrote:
> Mathieu Malaterre wrote:
> > jpr@creatis.univ-lyon1.fr wrote:
> >
> > > "12\10", should work, too, right?

>
> Yes, of course.
>
> > > How can I deduce that the Spacing is 0.3 * 0.25, and not 0.6*0.5?

>
> You can't. The ratio is less information than the pair of pixel
> spacings.
>
> >
> > After browsing the archive, the only answer that actually make sense is
> > one made by D. Clunie:
> >
> > ...
> > I think it is also fairly unusual for viewers and displays
> > to handle non-square pixels "properly" anyway.
> > ...

>
> That's a bit pessimistic! You can easily get non-square pixels by
> doing MPR; imagine an axial CT with 0.5 mm square pixels and 1mm slice
> separation -- the orthogonal views will have 0.5 mm by 1 mm pixels.
> Our viewer handles them fine. I'd be shocked if other commercial
> viewers did not.
>
> -Steve

5. ## Re: Pixel Aspect Ration vs Pixel Spacing

> ...
> I think it is also fairly unusual for viewers and displays
> to handle non-square pixels "properly" anyway.
> ...

I would just add that I think it is very poor engineering (an
understatement..., I could have been more blunt) to NOT handle it
properly. It is just a matter of interpolation to make them square. I
just don't want to get people the impression that it is OK not to
display these images correctly.

Herman O.

dmcnamara@vitalimages.com wrote:
> As a matter of fact, our customers have had numerous occasions where
> non-square pixels were not being displayed correctly. Specifically,
> the other vendors were only using Pixel Aspect Ratio, which works for
> most of the IODs out there. But, CT/MR/PET... use Pixel Spacing, not
> Pixel Aspect Ratio, which means that the images that were displayed
> looked "squished"... Ultimately, we were forced to "fix" our images,
> and make our pixels square, to satisfy the customer complaints.
>
> Steve Robbins wrote:
> > Mathieu Malaterre wrote:
> > > jpr@creatis.univ-lyon1.fr wrote:
> > >
> > > > "12\10", should work, too, right?

> >
> > Yes, of course.
> >
> > > > How can I deduce that the Spacing is 0.3 * 0.25, and not 0.6*0.5?

> >
> > You can't. The ratio is less information than the pair of pixel
> > spacings.
> >
> > >
> > > After browsing the archive, the only answer that actually make sense is
> > > one made by D. Clunie:
> > > http://groups.google.com/group/comp....b4ccdc2?hl=en&
> > >
> > > ...
> > > I think it is also fairly unusual for viewers and displays
> > > to handle non-square pixels "properly" anyway.
> > > ...

> >
> > That's a bit pessimistic! You can easily get non-square pixels by
> > doing MPR; imagine an axial CT with 0.5 mm square pixels and 1mm slice
> > separation -- the orthogonal views will have 0.5 mm by 1 mm pixels.
> > Our viewer handles them fine. I'd be shocked if other commercial
> > viewers did not.
> >
> > -Steve