We are happy to announce the 1st Secure and Privacy-Preserving Machine Learning for Medical Imaging Workshop and Tutorial co-organized by the Technical University of Munich, Imperial College London and OpenMined. The even will consist of a half-day workshop with research paper presentations and a half-day tutorial which will introduce participants to applied privacy-preserving machine learning tools. The event will take place on the 27th of September or the 1st of October, 2021.
Participants are NOT required to take part in both the workshop and the tutorial (but are encouraged to).
We are inviting author teams to submit their papers by June 27th, 2021 (01:00 AM, Pacific Time). Accepted papers will be assigned either an oral presentation or a poster presentation. As it is yet unclear whether the conference will take place on-site or virtually, presentations and posters may be virtual.
Authors will be notified of acceptance or rejection on the 19th of July, 2021.
Authors of accepted submissions will be invited to submit the camera-ready versions including any revisions by August 1st, 2021 (05:00 PM, Pacific Time).
There will be workshop proceedings, published with the official MICCAI 2021 proceedings (LNCS). Concurrent/dual submissions are therefore NOT allowed. Review will be double-blinded.
Papers should be submitted anonymously using the provided template. Manuscripts are limited to 8 pages in length. References and supplementary material/appendices do not count towards this limit. The overall length of the paper should not exceed 12 pages. The MICCAI template may not otherwise be modified. Submissions will be handled by the CMT website.
The workshop encourages submissions from all areas of privacy-preserving machine learning in the medical imaging domain, including (but not limited to):
We invite submissions in the following tracks:
We will honour the best submission with a best paper award.
We are determined to make the workshop inclusive, accessible and diverse. Please contact us if you require any assistance in advance. By participating, you are required to conform to the MICCAI code of conduct. We particularly encourage submissions by underrepresented groups in machine learning and will honour the best submission by an underrepresented group and provide the opportunity for mentoring by the OpenMined mentoring team.
If you would like to become a reviewer, please contact the workshop organisers directly.
|08:00 - 08:30||Introduction to PPML with PriMIA and theoretical background.|
|08:30 - 10:00||Live coding: example using PriMIA|
|10:00 - 10:15||Break|
|10:15 - 11:00||Intro to differential privacy|
|11:00 - 12:00||Live coding: Privacy-preserving deep learning using deepee|
The tutorial is open to participants at all levels of experience and explicitly designed for beginners in PPML. Participants are, however, expected to be familiar with training deep neural networks in computer vision tasks.
Due to the high probability of the event being online, we expect participants to bring their own computer. A GPU is explicitly not required, as we will be using datasets suitable for CPU training.
In preparation for the tutorial, we kindly ask participants to install the PriMIA and deepee libraries in two separate virtual environments. We strongly encourage the utilisation of the conda package manager.
In preparation for the tutorial, participants are encouraged to visit https://courses.openmined.org and https://www.udacity.com/course/secure-and-private-ai–ud185 to learn the basics of privacy-preserving machine learning using free courses offered by OpenMined.
The concrete outline of topics that will be covered is as follows:
The tutorial will be framed by short-form lectures introducing the pertinent topics.
Participants are able to take advantage of mentoring by the OpenMined team by joining the OpenMined Slack Channel.