XW2021 Round Table: Supporting Machine Learning with XNAT

Time: Day 1, 11:15 – 11:45 am CDT

Host: Marc Modat, Kings College London.

Participants: Aaron Mintz (Washington University School of Medicine), Jorge Cardoso (Kings College London), Ahmed Harouni (NVIDIA Clara)

Questions and Answers

When possible and as time permitted, questions that were brought up in the Q&A module during each talk were addressed in real time by the presenter. Other responses were entered in the Q&A interface itself. Those written responses are included below.

Questions	Answers
During the xnat upload, is there a way to anonymize using predefined whitelists and eliminating all unknown datatypes?	From Dave Maffitt: "For private tags, the anonymization scripting language has a retainPrivateTags function. There is not corresponding function for normal tags."
Not related to the ML topic, but some XSync questions. Is there any suggestion for syncing the shared subjects/experiments? Also, is there a way to bypass the firewall - having trouble connecting to the another (https://) xnat website	We will respond to this later!
Many of the XNAT institutional and federated models are using multiple, tiered XNAT structures, how are the XNAT to XNAT transfers or sync managed, is it all done with xsync or is there something else that is used ?	We recommend that you attend XNAT Federation Town Hall session tomorrow for an update the panalists will be able to answer your questions in detail.
Is DQR generally recommended over CTP as we were going to set up CTP on a VM but open to DQR if that is more robust	Ajay, I think this is worthy of a breakout conversation in Spatial Chat tomorrow. I believe both are valid and valuable approaches, but it depends on your use case. In terms of robustness, DQR certainly offers more fine-grained control on importing and relabeling within the XNAT UI.
Sorry, are ML Panelists available for questions tomorrow in the poster sessions?	Yes, and there will be a poster on TIP as well