Secondary XNAT or dcm4chee as local PACS ?

158 views
Skip to first unread message

Michael Dayan

unread,
May 10, 2022, 10:33:03 AM5/10/22
to xnat_discussion
Hi Everyone,

We are in the process of trying to identify the best solution to manage MRI (and EEG) data, with a PACS for local raw data (potentially transferred to hospitals) and another remote database for de-identified data (for research).

For the remote de-identified research database we already set our choice on XNAT, but for the raw PACS data we are hesitating between different solutions:
1) dcm4chee (then transfer de-identified data to remote XNAT or identified data to hospital, i.e. dcm4chee -> XNAT remote or hospital)
2) dcm4chee and a local instance of XNAT (dcm4chee -> XNAT local --> XNAT remote or hospital)
3) local instance of XNAT to do everything as a local PACS and completely replace dcm4chee (XNAT local --> XNAT remote or hospital )

It would be great to have feedback on these 3 possibilities.

If that help our current train of thoughts for solution #2 were:
- would there be a duplication of data (twice the storage) or could the local XNAT point to data stored on the local dcm4chee ?
- would a local XNAT really bring more useful features compare to dcm4chee only (knowing that it will be the remote XNAT that would be used for research) ? 

And for solution #3:
- could XNAT do everything that dcm4chee does, and in a robust way (risk of more functionalities from XNAT causing more potential failures ?) ?
- does XNAT offer more control on the raw data access and more audit logs, or does dcm4chee do that very well ?

Again, any feedback would be very much appreciated.

Best,

Michael

Daniel Drucker

unread,
May 20, 2022, 11:28:48 AM5/20/22
to xnat_discussion
We went through this same decisionmaking and opted for using Orthanc for the raw data. 

We treat the Orthanc server as an immutable archive and Source of Truth.  Only MR techs and IT staff have access to it. XNAT by contrast is what researchers get access to. On the XNAT side, researchers can edit/delete scans, etc... but if they mess up, it's always possible to push the raw data from Orthanc to XNAT again.

Daniel Drucker
McLean Hospital Imaging Center


Michael Dayan

unread,
May 25, 2022, 7:09:15 AM5/25/22
to xnat_di...@googlegroups.com

This is great to know, many thanks for the feedback Dr Drucker. Do you mind if I ask you a few questions ?

  • Any specific reasons for choosing Orthanc instead of say dcm4chee ?
  • Why did you decide not to use an independent XNAT instance to replace your PACS ?
  • Would you still choose the same solution today if you had the choice or would you suggest another one ?

--
You received this message because you are subscribed to a topic in the Google Groups "xnat_discussion" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/xnat_discussion/6Xuh_Evu11Q/unsubscribe.
To unsubscribe from this group and all its topics, send an email to xnat_discussi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/xnat_discussion/84ad3545-1507-4e70-8fba-032c31b04206n%40googlegroups.com.

Daniel Drucker

unread,
May 25, 2022, 8:25:49 AM5/25/22
to xnat_discussion
The answer to the first two is basically the same - simplicity. For us, lack of features was a feature. All we wanted was a drop-in solution that we could absolutely rely on to store the unmodified dicoms. dcm4chee and XNAT both are "hairy" - they are all things to all people, and that's good for some use cases if your requirements are complex, but if all you want to do is have a system-of-record archive, I don't want all that complexity.
And I use the Docker version of Orthanc, using docker-compose to build it alongside Postgres (and an apache front-end that provides an Okta SSO login), so that makes life even easier! 

I would absolutely choose Orthanc for this again. It's been delightful to maintain (which is to say, I never have to touch it!) and use.

--
Daniel M. Drucker, Ph.D.
Director of IT
McLean Hospital Imaging Center
A Harvard Medical School Affiliate

Michael Dayan

unread,
May 25, 2022, 8:38:43 AM5/25/22
to Daniel Drucker, xnat_discussion
Thanks a lot for the feedback. Dcm4chee also proposes some microservices architectures now and I like this approach as well. I'll check out Orthanc, thank you!



From: xnat_di...@googlegroups.com <xnat_di...@googlegroups.com> on behalf of Daniel Drucker <ddru...@mclean.harvard.edu>
Sent: Wednesday, May 25, 2022 14:25
To: xnat_discussion <xnat_di...@googlegroups.com>
Subject: Re: [XNAT Discussion] Re: Secondary XNAT or dcm4chee as local PACS ?
 

CORADS reports

unread,
Jun 1, 2022, 5:21:26 AM6/1/22
to xnat_discussion
That's a good suggestion.
How do you anonymize the data for researchers? How to they segregation it based on pathologies. Because the raw data from scanner wouldn't be radiologist read data. One way work around is research team having their own radiologists but that's expensive for some setups. 

Thanks in advance. 
Reply all
Reply to author
Forward
0 new messages