SOP Instance UID changing

[Bartosz Wiklak]

Hi,

Consider a scenario, when implicit little endian image is burned on
optical media. It is required to convert such image to Explicit little
endian. Should such new image have the same SOP Instance UID or it has
to be changed? What about series/study and f.e. instance number?

What about changing Implementation Version Name or adding Icon Image
Sequence?

In the David Clunie's "DICOM Structured Reporting", chapter 8 "Copies
and Versions", page 191 it is said that:
"DICOM image objects are defined such that if the pixel data of a
derived image is different
from the pixel data of a source image, and this difference is expected
to affect
professional interpretation of the image, the derived image shall have
a different UID."

I believe that these situations do not "affect professional
interpretation of the image" but it's not logical but some kind of
legal statement (if you know what I mean - there is some place for
interpretation).

What do you think?

Bartek

[mathieu]

On Jun 26, 10:12 am, Bartosz Wiklak <bwik...@gmail.com> wrote:
> Hi,
>
> Consider a scenario, when implicit little endian image is burned on
> optical media. It is required to convert such image to Explicit little
> endian. Should such new image have the same SOP Instance UID or it has
> to be changed? What about series/study and f.e. instance number?

You are not changing anything in the DICOM dataset, only it's
representation on disk, so: no you do not (and should not) change
anything.

> What about changing Implementation Version Name or adding Icon Image
> Sequence?

Group 0002 still does not impact DICOM dataset so you are fine.

SOP Instance UID does not refer to Icon Image, so -again- no you do
not need to change it. However please note the usual recommendation:
if you previously send this instance uid to a PACS, sending it again
(with the icon this time) you are not garanteed that the new instance
on the PACS will have the Icon Image Sequence.

> In the David Clunie's "DICOM Structured Reporting", chapter 8 "Copies
> and Versions", page 191  it is said that:
> "DICOM image objects are defined such that if the pixel data of a
> derived image is different
> from the pixel data of a source image, and this difference is expected
> to affect
> professional interpretation of the image, the derived image shall have
> a different UID."

Lossy compression would be applicable in this case (also explicit
transfer syntax).

-Mathieu

[Bartosz Wiklak]

Thank you very much Mathieu, it's a very comprehensive answer. I did
expect this but it is good to be sure.