Computer Vision News - January 2019

17 ABCD Neurocognitive Prediction Challenge or both? Kilian: While it’s not tested for in the challenge, it would be really nice if people don’t just apply approaches to do prediction on the score, but also to let us know how they came to those results. What changes in anatomy cause to drive the prediction of the fluid intelligence scores? That would be really helpful for ABCD and, having said that, there is no way of testing for that in the challenge so we didn’t attempt to do that, but it gives additional brownie points. Wes: We’re not observing changes in brain matter. We’re observing differences in brain matter between children. This is the baseline data. As Susan said, this is a developmental study, a longitudinal study. We’ll be repeating brain scans on the same children multiple times over 10 years. In fact, every other year we’ll get another full brain scan with all the different modalities that Kilian mentioned before. We’d like to, hopefully, depending on how well- received this initial challenge is, to do future challenges where we actually do individual-level predictions. That remains to be seen how we actualize that, but then we would use longitudinal imaging data on kids to predict their future values on some quantity, maybe not neurocognition. Maybe it has something to do with psychopathology or something else. Going forward, we’d like to continue these contests using the longitudinal data and see how well we can predict how trajectories change within a person. Kilian: We added a lot of information on our website. We will also make available the scoring algorithm, every component of it. If people have any questions, that’s something to go on to. It takes a little bit of time to get access to the data because NDA (National Institute of Mental Health Data Archive) requires a sign-off by the institution, not by the participant, but by the institution of the participant to get access to it. People should start early. This can take several weeks. Often the institutions that people are affiliated with take a while to fill out these forms. Keep that in mind when trying to participate in the challenge. There is also an email list people can sign up to and a little tutorial on how to apply for access to NDA. I hope we’ll see a lot of the people at MICCAI 2019 in October when we are organizing a workshop about the results of the challenge. Susan: We’re really interested to see what kind of algorithms we generate. This is really exciting data. Good luck! Challenge Computer Vision News Wes Thompson is Associate Professor of Biostatistics at the University of California in San Diego and the Associate Director of the DataAnalysis and Informatics Core for the Adolescent Brain and Cognitive Development Study, in charge of consortium and biostatistics