Anonymize dicom before sending to AI

General
1

Hello everyone,
I want to anonymize the dicom before send to AI companies for AI generated report in form of dicom instance. When they send back the result that I want to store in the original dicom.
Or is there any way to de-anonymization the dicom to get it’s original form ?
Or is there any other way to perform this kind of operation ?
Any suggestion or help will be a great .
with regards
Ankush
Have a great day to all .

2

Hello, Ankush,

There is no “magic button” doing exactly what you described.
That being said, all the tools to achieve your goal are available.

You could use the API of Orthanc to modify the instances (Anonymization and modification — Orthanc Book documentation) before sending them to the AI company and then use it again to restore the data.
Of course, you will have to keep the information in a buffer or something like that to be able to restore it.
That could be done with a dedicated database, or with an extra Orthanc…

HTH,

3

Hi Ankush,

Note that you can possibly use the AnonymizedFrom metadata to rematch incoming data.

HTH,

Alain

4

hey @bcrickboom ,
thankyou for suggestion and resources .

5

hey Alain,
thankyou for suggestion. I am almost using the same .

7

A possible way to merge the incoming report in the original study is to use the /merge route and merge the newly received series/instances in the original study, than you can track using the AnonymizedFrom tag, as suggested by Alain.

Initially, you have an OriginalStudy, OriginalSeries and OriginalInstances. Then you have AnonStudy, AnonSeries and AnonInstances.

AnonInstances gets send to AI and you receive AnonReportInstances. Hopefully, this report should still still be part of AnonStudy

So, from AnonReportInstances, you can find OriginalStudy with the medata, and use the /studies/<ORIGINALSTUDY>/merge route with your AnonReportInstances that would then, after the merge, be part of the original study.

That’s how I would solve this problem.

Hope this helps

8

hey @benjamin.golinvaux ,
Thankyou so much for explaining the entire process in such a way.
I am using the same to achieve my goal.
With regards
Ankush.

9

My pleasure ! Let us know if you run into any problem.

10

Definitely , I will sir.