Hi everybody,

we had some problems de-identifying studies with Grayscale Softcopy Presentation States:
PS3.15 Table E.1-1. lists the Graphic Annotation SQ (0070,0001) as "D" (Basic Profile).
"D" subsequently applies to all attributes in the SQ.
Doesn't this necessarily result in loosing the referenced SOP Instance in Referenced Image Sequence (0008,1140)?

Thanks,
Silvia

Le mercredi 11 août 2021 à 11:45:27 UTC+2, Silvia Winkler a écrit :
> Hi everybody,
>
> we had some problems de-identifying studies with Grayscale Softcopy Presentation States:
> PS3.15 Table E.1-1. lists the Graphic Annotation SQ (0070,0001) as "D" (Basic Profile).
> "D" subsequently applies to all attributes in the SQ.
> Doesn't this necessarily result in loosing the referenced SOP Instance in Referenced Image Sequence (0008,1140)?
>
> Thanks,
> Silvia
Hi,

I would assume that the action code is applicable to the content of the sequence, unless the content has its own action code in the list. It's not explicitly written but that's my understanding.
In that case, the SOP Instance UID (0008,1155) has the action code "U" which means UID relationships should be preserved.

best
Thomas

Thank you, Thomas !
Your approach seams reasonable - but wouldn't that result in replacing (0008,1150), i.e. the Referenced SOP Class UID, due to ('D')?

Best,
Silvia

Le lundi 16 août 2021 à 14:53:54 UTC+2, Silvia Winkler a écrit :
> Thank you, Thomas !
> Your approach seams reasonable - but wouldn't that result in replacing (0008,1150), i.e. the Referenced SOP Class UID, due to ('D')?
>
> Best,
> Silvia
If you apply the standard very strictly, yes.

But I would argue that the standard gives guidelines, not strict rules, about de-identification. In the text, there are multiple caveats about how these guidelines may not apply to every context. So the general approach is to follow the guidelines where it's reasonable, and apply some common sense and good justification when it's not. This is what I did when we designed the de-identification process of our upload system for clinical trials. As long as you can justify each decision, I think you're fine.

In particular, SOP Class UIDs seem to me very safe attribute to keep without change, imo.

best
Thomas

I agree, the SOP Class UID is very safe to keep :)
But the point is, that de-identification is hard to implement, if the rules resulting from the standard are not strict.

In case of sequences, table E.1-1 in PS3.15 requests different actions, and for me it is not clear how to interpret this.

In case of PR-Objects, there are relevant sequences to make the PR work for the study are:
(actions according Basic Profile)
(0008,1115) Referenced Series Sequence is not listed --> No action, but check the content
> (0008,1140) Referenced Image Sequence is X/Z/U* --> action
>> (0008,1150) Referenced SOP Class UID is not listed --> keep unchanged
>> (0008,1155) Referenced SOP Instance UID is U --> apply the 'U'
...
(0028,3110) Softcopy VOI LUT Sequence --> is not listed --> No action, but check the content
> (0008,1140) Referenced Image Sequence is X/Z/U* --> action
>> (0008,1150) Referenced SOP Class UID is not listed --> keep unchanged
>> (0008,1155) Referenced SOP Instance UID is U --> apply the 'U'
...
(0070,0001) Graphic Annotation Sequence --> D --> action
In that case, it is not clear for me, how to apply the 'D'
* apply 'D' to all attributes in the sequence
--> completely loose the graphic annotation

* apply 'D' to all attributes in the sequence, that are listed to be relevant (i.e. having an action defined)
--> apply D to the Referenced Image Sequence
--> apply D to the Referenced SOP Instance UID --> loose the reference
* apply 'D' to all attributes in the sequence, that are listed to be relevant, but apply THEIR action
--> would result in same behavior as described above for X/Z/U*
So the question is:
What is the intended effect of the D for Graphic Annotation SQ (instead of having X/Z/U*)?
Probably the deletion of the graphic annotation is the intended effect, because the annotation maybe could contain identifying information.

On Wednesday, August 11, 2021 at 11:45:27 AM UTC+2, Silvia Winkler wrote:
> Hi everybody,
>
> we had some problems de-identifying studies with Grayscale Softcopy Presentation States:
> PS3.15 Table E.1-1. lists the Graphic Annotation SQ (0070,0001) as "D" (Basic Profile).
> "D" subsequently applies to all attributes in the SQ.
> Doesn't this necessarily result in loosing the referenced SOP Instance in Referenced Image Sequence (0008,1140)?
>
> Thanks,
> Silvia

Hi Silvia,
With deleting the sequence you will lose the embedded Referenced Image Sequence as all items in the sequences are deleted.
Keeping the reference would not make sense either as there is no longer an annotation to be shown on the images.

Deletion is ok as the reference(s) only applies for the annotation specifically as the text defines the content as:
"The subset of images and frames listed in the Presentation State Relationship Module, to which this graphic annotation applies."
And it will only have content if a subset of the images is referenced otherwise it is applicable for all images referenced.
So the full set of references is still present in the Presentation State Relationship Module.
With that you don't loose any relations.

> Le mercredi 11 août 2021 à 11:45:27 UTC+2, :
> > Hi everybody,
> >
> > we had some problems de-identifying studies with Grayscale Softcopy Presentation States:
> > PS3.15 Table E.1-1. lists the Graphic Annotation SQ (0070,0001) as "D" (Basic Profile).
> > "D" subsequently applies to all attributes in the SQ.
> > Doesn't this necessarily result in loosing the referenced SOP Instance in Referenced Image Sequence (0008,1140)?
> >
> > Thanks,
> > Silvia
> Hi,
>
> I would assume that the action code is applicable to the content of the sequence, unless the content has its own action code in the list. It's not explicitly written but that's my understanding.
> In that case, the SOP Instance UID (0008,1155) has the action code "U" which means UID relationships should be preserved.
>
> best
> Thomas

Hi Thomas,

You are right about the rule for the SOP Instance UID, but in this case the D for the sequence removes the SOP Instance UID already and with that there is nothing left to apply the U on.
So a D for a sequence is removing the full sequence and with that all its items.
In this case the main reason is as Silvia correctly said the annotation might have personal information encoded.

Regards,
Wim

Hi Wim,

thank you for your explainations and for clarifying that graphic annotations should be deleted.
But shouldn't in this case the option simply be 'X' ?

'D' requires a replacement with a non-zero length value consistent with the VR.
I cannot imagine any implementation for this.

Best,
Silvia