Analytic Imaging Diagnostics Arena (AIDA)

Researchers around a computer. John Sandlun

AIDA is a national arena for research and innovation in medical image analysis. AIDA is a cross-disciplinary collaboration aiming for largescale use of AI in healthcare. In the arena, academia, healthcare and industry meet to translate AI technology advances into patient benefit in the form of clinically useful tools. CMIV is the host and physical meeting place of AIDA but aims to assist all Swedish actors in this domain.

The technical development within artificial Intelligence (AI) has been extremely strong in recent years. Modern AI is a toolbox that fits perfectly into the healthcare vision of “precision medicine”, the fully tailored treatment for each patient. However, very few modern AI solutions have yet reached actual use in imaging diagnostics. The reason is that the step from experiments to clinical routine entails many challenges. Even the most powerful algorithms need to be carefully placed in a context of workflow and interaction for the innovations to be useful.

A National Arena

Analytic Imaging Diagnostics Arena (AIDA) is a national arena for research and innovation in medical image analysis. AIDA’s objective is to develop AI-based decision support solutions for imaging diagnostics that reach all the way to clinical use. An underpinning fundamental insight is that this complex challenge requires both interdisciplinary and cross-sectoral collaboration.

Claes Lundström is a Professor in medical visualization and the director of AIDA.

-We saw a need for more knowledge about how to translate AI innovations into actual use in clinical routine, on the healthcare side as well as in technical research. Without more knowledge about each other the engineers will continue to produce solutions they believe are beneficial but in reality, never reach the patients and the healthcare professionals will continue not knowing what to ask for to refine the innovations, Claes explains.

Financial Support as well as a Place to Meet

Most of AIDA’s resources are used for projects developing AI-based decision support solutions. These are run by research groups in industry and academia across Sweden, in collaboration with healthcare providers. AIDA also offers a core environment at CMIV, with technical infrastructure designed to support the development projects. Perhaps even more important is the meeting place aspect of the core environment, where workshops and meet-ups are frequently organized, providing valuable knowledge and exchanges. To give healthcare the right knowledge base to drive the AI development in the most effective direction, AIDA offers both clinical and technical fellowships where care provider employees or engineers carry out an individual project as continued education.

-The goal is to introduce mechanisms that will continue to build valuable collaborations and innovations even after the AIDA project is concluded, Claes continues.

Large-Scale AI Computations

Developing AI innovations requires large amounts of relevant data and the relevant data is found in healthcare. There is an extensive uncertainty about how to share data in a secure, ethical and legal way. AIDA has started up a Data Hub in order to facilitate data sharing in different ways.

-We collect data sets that can be shared for important research and we try to prioritize what the healthcare actually needs help with. If you look at the typical research done in AI right now it is done with data that just happens to be available, not necessarily prioritizing what the healthcare needs help with, Claes says.A window from the data hub.A data set entry in the AIDA Data Hub.Images showing histology tissue from breast cancer axillary lymph nodes. from

Before starting the data collection, it was necessary to work through the legal and ethical questions.

-Several laws are involved here so for us it was important to state how to work with this kind of data. The thorough effort done has resulted in a data policy that is publicly available to guide anyone who work with these questions.

Working with large data sets requires an unusual amount of computer power that is not possible for most research groups to host. AIDA is now investing in a unique system with 16 GPU processors that can handle large-scale AI computations. The system will be shared between the AIDA researchers across the country and will enable AI model training that is faster and includes larger data volumes than was possible before.

Around 20 innovation projects are currently running or has been concluded under the AIDA flag. The projects are initiated by both technical and medical researchers from both academy and industry. The knowledge about AI has increased among pathologists and radiologists in Sweden and the departments are more ready to embrace AI as part of their profession now, than a few years back.

AIDA is an initiative within the Strategic innovation program Medtech4Health, jointly supported by VINNOVA, Formas and the Swedish Energy Agency. 

Engaging with AIDA

All companies, academic groups and health care providers interested in AI in diagnostic imaging are welcome to join AIDA. There are many possibilities to formal AIDA efforts with partial funding. But you can join also without defined project ideas as a Network partner (full description here), and among other benefits take part of the knowledge exchange and shared data sources within AIDA.

You can read more about how to apply for projects and fellowships on the Medtech4health site

Key Publications

Annotations, Ontologies, and Whole Slide Images - Development of an Annotated Ontology-Driven Whole Slide Image Library of Normal and Abnormal Human Tissue.
 Lindman K, Rose JF, Lindvall M, Lundström C, Treanor D. J Pathol Inform. 2019;10:22. Published 2019 Jul 23. doi: 10.4103/jpi.jpi_81_18
AIDA data sharing policy

Report No. LiU-2017-01901, Linköping University, 2019. 

Axillary lymph nodes in breast cancer cases

Sofia Jarkman, Martin Lindvall, Joel Hedlund, Darren Treanor, Claes Lundstrom, and Jeroen van der Laak (2019) doi:10.23698/aida/brln


Research at CMIV