As the landscape of medical machine learning continues to evolve, access to diverse, representative, and inclusive datasets has become paramount for addressing global healthcare challenges effectively.
The MICCAI 2024 conference is featuring the first Open Data Session as part of the Open Data Initiative which aims to empower collaboration and innovation by facilitating and promoting sharing of medical imaging datasets. The session will run in the afternoon of October 7, in parallel with the main conference track, from 13.30-18.00 in the Diamont room.
"The response to this new initiative has been very positive; everyone we speak to acknowledges the value of Open Data, with some people being surprised that this is the first time one is being organized,” said Martijn Starmans, Co-Chair.
"Besides the positive recognition of the movement overall, we see people also have a need for help in achieving open data. We hope in the future the initiative can continue to promote open data, and bring more practical guidance for researchers to ensure the published data is FAIR and of good quality.”
The Open Data Session will feature seven diverse oral presentations, each about a newly published open dataset, with just over half presenting African data. These were selected from paper submissions describing novel datasets, especially from the African continent and other underrepresented populations and diseases - including methodologies for the data collection and curation, and innovative approaches for utilizing them in medical imaging research. While most of the data is from Africa, there are also datasets from under-represented countries in the open data domain, such as India.
"We also saw a trend in the type of authors present,” said Apostolia Tsirikoglou, Co-Chair. ” While often led by researchers from their native countries, there was often co-authorship by several partners from other continents, mainly Europe, or double affiliations with both African and European institutes. This shows that there is support from the community to help those who have fewer experience to make data publicly available, which is great.”
The Open Data Session will also feature presentations from two keynote speakers.
Udunna Anazodo, PhD, Assistant Professor and member of the Montreal Neurological Institute (The Neuro) at McGill University. She will be highlighting the importance of publishing open data from Africa and sharing her experience with publishing open data and the hurdles she overcame.
Fred Prior, PhD, is the Distinguished Professor and Chair of the Department of Biomedical Informatics and Professor of Radiology at the University of Arkansas. He is also the lead of one the biggest open data initiatives and medical imaging repositories, The Cancer Imaging Archive (TCIA). He will showcase the TCIA and discuss more about the technical aspects of how to anonymize data, curate data, and manage a repository.
As part of this year's initiative, Tsirikoglou and Starmans are guest editors of a special issue of the MELBA journal, where all accepted papers will be published. It is expected that the special issue will be launched shortly after the conference. Besides all accepted papers, submissions to the special issue will stay open to allow other researchers to submit.
"We hope other researchers will feel inspired by our session and the showcased MICCAI 2024 Open Data datasets to publish open data on their own,” said Starmans. "Researchers who have any questions about this, do not hesitate to contact us.”