We offer leading edge services to researchers in Swedish medical imaging AI, in collaboration with the Analytic Imaging Diagnostics Arena (AIDA), the Center for Medical Image Science and Visualization (CMIV), and the SciLifeLab Bioinformatics Platform and Data Center.

One example is the AIDA DGX-2 Service which is currently the largest national GPU resource for processing sensitive personal data for research in Sweden. Further exampes are introduced in overview sections below.

These services are offered to AIDA partners. All Swedish organizations active in medical imaging diagnostics or AI are invited to become a member of AIDA.

AIDA DGX-2 Service

AIDA Data Hub is collaborating with Nvidia to operate a DGX-2 machine learning system at CMIV, as a service for leading edge AI researchers in Swedish medical imaging diagnostics, secure enough for use with sensitive personal data. We believe that these experts will produce the best results if given the freedom to exercise their expertise according to their best judgement.

Nvidia DGX-2 Machine learning system

The service was established in a phased approach, opening for research use immediately on its delivery day of 2020-03-05, and from then on adding further features, stability, and guarantees in stages. The final step to full production was taken on 2020-11-09, when it was approved for use with sensitive personal data.

The service is set up to conform with Linköping University information security guidelines for systems intended for processing information of extreme confidentiality, in collaboration with Linköping University Data Protection Officers. System security was validated using international penetration testing expertise.

Service is offered to AIDA partners, in compute projects, based on Data Processing Agreements, or Servcice Agreements for anonymous data. Project members get full access to private project VMs and persistent storage through personal VPN accounts, and use secure SSH pubkeys to SSH to VMs and SFTP to storage. Outgoing connections are limited and appoved based on feasibility, where IP:PORT is easiest, and distributed services over CDN are harder. Storage/VMs are granted based on availability (cf booking sheet). Projects work with own private data. AIDA Data Hub Datasets are pre-mounted on request. Projects cannot access resources belonging to other projects.

VMs run latest v4.x Nvidia DGX OS, have multiple Nvidia V100 GPUs on PCI passthrough, and mount project storage over 10Gbit connection with FSC to local NVMe raid.

More information on the AIDA internal wiki DGX-2 page.

AIDA Nextcloud

AIDA nextcloud

AIDA Data hub uses nextcloud.aida.scilifelab.se to allow AIDA Partners to share anonymous research data with the world. Offered in collaboration with the SciLifeLab Data Center.

For sharing non-anonymous data, please see the AIDA DGX-2 Service above.

More information on the AIDA internal wiki Nexcloud page.

Sharing large amounts of data

We recommend using the nextcloud client, or rclone.org with the WebDAV interface. Always use the latest version. Both nextcloud and rclone develop fast.

AIDA users can access their data through their WebDAV url, on the form:

https://nextcloud.aida.scilifelab.se/remote.php/dav/files/USERNAME/

For share links, use the url:

https://nextcloud.aida.scilifelab.se/public.php/webdav

The USERNAME is then the last part of the link, i.e:

https://nextcloud.aida.scilifelab.se/s/USERNAME

AIDA PACS

Picture Archive and Communication System

AIDA offers access to a cloud-based production grade Picture Archive and Communication System (PACS) with both a desktop client (Windows) and a web client for viewing and interacting with image data. The AIDA PACS can be used by AIDA Partners to explore different integrationopportunities, annotate data and perform reader studies. Please reach out to AIDA for access and further information. Provided by AIDA.

AIDA Wiki

AIDA uses an AIDA Confluence space as an internal wiki, for documentation and internal materials. Offered in collaboration with SciLifeLab.

AIDA partners get accounts on request. Click the link, or remember to provide name, affiliation and email address in your request!

AIDA GitHub

AIDA uses an AIDA Team in the SciLifeLab Bioinformatics Platform (NBIS) GitHub account to facilitate code sharing and co-development in AIDA. Members can create public or private repos to admin themselves, and easily share with AIDA, NBIS, invited external parties, nobody, or the world.

Examples of AIDA-internal repositories include training materials from AIDA coding workshops, and PACS integration examples in pathology and radiology. Because of size limitations, Some large AI training data from AIDA coding workshops are made available on AIDA nextcloud instead.

More information on the AIDA internal wiki Nexcloud page.

AIDA Slack

The primary channel to all-to-all support and discussion in AIDA. AIDA uses SciLifeLab Bioinformatics Platform (NBIS) Slack:

More information on the AIDA internal wiki Chat page.