A Ultrathon is currently being designed – a 6-week long competitive pre-event hackathon addressing important medical AI questions from real clinical environments.

For more info and sign-up for the Ultrathon: Click Here

Application instructions

If you wish to attend our conference, you will need to apply and submit an abstract in our application system NORMA. If you have not used NORMA before, click REGISTER to create an account. When you enter the system, you can find the conference on the list of calls.

You can find the call by clicking on “Copenhagen Bioscience” in the top bar.

Evaluation of incoming applications: End of April 2021

Notifications to applicants: End of April 2021

Conference poster design by Anncharlott Berglar, PhD, Scientific & Medical Visualisation (SciVisLab), Web, LinkedIn, Instagram, Facebook.

Content: The main goal of the conference is to connect participants in authoring a whitepaper on the current global state of medical AI with recommendations and guidelines to implement it in the clinic. We are therefore interested in any abstracts that promote the discussion of medical AI and in particular its implementation. Content may range from basic and applied research, through to presentation of real-world examples or thoughtful identification and discussion of needs and the future of medical AI. We are particularly interested in submissions that highlight the reality of AI in medicine and medical research in contrast to the hyperbole surrounding the topic. We therefore invite all stakeholders for whom AI impacts in any way to submit their work for consideration. We hope to collect a diverse array of speakers to focus on the major issues at hand and avoid the technical jargon. We have therefore created several topics designed to overlap and cover the broad issues surrounding this subject. When submitting your abstract, please consider clearly identifying one or more topics your work might be considered for.

Novel data driven treatments: What does “data driven” mean and how is it different to what we have always done? What has changed to enable this, what does it take to employ it, and when and how will we see the impact of these new methods? Are we seeing such treatments in the clinic already and what were some of the challenges faced?

Governance and policy: Often maligned but universally agreed as essential. Medicine and AI are also both abundant in ambiguity and combining these three topics seems doomed to fail. We have therefore doubled the number of talks on this important topic. What is the policy supporting medical AI? What does the global governance landscape look like? At what stage of an mAI project does ethics fit in? What are we talking about when it comes to bias in models?

New collaborative models: Machine learning an AI has contributed toward significant gains in many industries. Those industries have contributed many methods back to data science. Their methods of collaborating and sharing data far exceed those in medicine. What are our barriers and how might we overcome them?

Real-time medicine: Data driven research requires a lot of data from many instruments and real-time medicine, for example by mobile health, could be both a major contributor to data collection and disseminator of personalized health advice. What does real-time medicine look like in practice and what major opportunities remain?

Merging data silos & omics: A major barrier to AI research and its broad implementation is the separation of datasets either intentionally or unintentionally through different interpretation of best practices in harmonisation, application of different ontologies, or mixed incentives. Omics technologies have vast promise but further exaggerate these disparities by contributing to one silo over another. Basic human research has mostly overcome these issues by centralizing omics data through public repositories, but how does clinical research fit into this landscape and how might hospitals in different locals best deploy data driven research from such resources?

Lifecycles of algorithms in medicine: Implementation of medical AI implies the same or better degrees of trust that we currently place in standard treatment. What does it, or would it, take for us to trust AI in a clinical setting? And once deployed, what would it take for such an AI solution/treatment to remain? What considerations would be needed to retire an algorithm? Current guidelines mostly classify AI in the “software as a medical device” category, but while this allows for its implementation does it promote trust and is it the appropriate translational vehicle to move AI from the laboratory to the bedside? If we consider the rigor of drug design and clinical trials, what are the analogs for mAI and what does phase-IV even look like?

Learning from COVID19: Before 2020, AI promised a lot to medical research and practice. But when COVID19 arrived much of the community fell back on tried-and-trusted methods. Of course, this was the correct move, but many who were not directly involved in the medical response also had the freedom to experiment with AI in diverse applications from triage to virology and epidemiology. What did they learn? What did first responders lack and what could AI have achieved if it were mature and trusted enough? Has the pandemic taught us anything about how we might do better? Or, was AI indeed a major player in our response and understanding of the virus?

Communication and leadership, stakeholder perspectives and success stories: Several catch-all topics for those who want to share how they / their group / their institute has handled AI in their setting, what the challenges were and how they overcame them. Fully implemented models with “Real World Evidence” on clinical impact can be submitted for this topic as well.

Lastly, we also accept research that focus on methods that address specific challenges and/or methods in precision medicine:

• State-of-the-art algorithms
• Interpretability and explainability
• Algorithm confidence and trust
• Robust validation approaches
• Addressing the domain gap
• Supervised learning for small cohorts
• Collaborative or competitive modelling on common benchmarks
• Generative models for simulating patient data
• Privacy preserving data sharing and federated learning
• Accessibility of algorithms for non-data scientists

If you are not able to upload an abstract, we also welcome that you upload a Motivation instead in which you explain why you should be included in the conference.