49 In the Picture: Medical Imaging Datasets Computer Vision News Computer Vision News Campello, the virtual sessions for the ‘In the Picture: Medical Imaging Datasets’ workshop organized by Amelia and Veronika,” was Camila’s comment. “The event brought together a wonderfully diverse group of attendees, ranging from computer science students to professionals with regulatory expertise. During the online sessions, we discussed the importance of dataset quality, noting that poor documentation and a lack of metadata often hinder external validation efforts across subgroups. Coming from a continual learning background, lack of metadata has been a significant obstacle in simulating data drift scenarios in clinical situations. We also highlighted the general need for clearer guidelines on data and model sharing, including practices like versioning. How we address these issues going forward, while protecting patient privacy, will determine the success and speed of radiology AI systems in the coming years." On the second day, we synced with Víctor to gather the insights of the online participants, allowing for a virtual contribution to the workshop. In addition to the scientific aspect, we chose a venue by the sea. This setting allowed attendees to enjoy Denmark's renowned cycling culture, with some even taking the opportunity for a refreshing swim! In the evening, Veronika organized a lively pub quiz, which participants really enjoyed. We are thrilled with the high level of engagement and interaction throughout the sessions. Thanks to everyone for joining! Veronika’s final session centered on strategies for building and maintaining communities, echoing our commitment to continuing the conversation beyond the workshop. We plan to keep the webinar Datasets through the LookingGlass series going, although in a more decentralized format. We would like to thank the Independent Research Council Denmark (DFF) and the Danish Data Science Academy (DDSA) for funding our workshop. Stay tuned for what’s coming next as we continue to explore the evolving landscape of medical imaging datasets!
RkJQdWJsaXNoZXIy NTc3NzU=