UCL researchers win ABCD-NP Challenge for MICCAI 2019

The challenge was to predict fluid intelligence scores using only the brain scans from 4,400 10-year-old children. Brain scans and intelligence scores from an additional 4,000 children were provided for training of predictive models.

The research team, led by Dr Neil Oxtoby and Prof Janaina Mourao-Miranda, included researchers from the UCL Centre for Medical Image Computing (CMIC), the Wellcome/EPSRC Centre for Interventional and Surgical Sciences (WEISS), and the Wellcome Trust Centre for Neuroimaging.

The team setup a hackathon to brainstorm and experiment with approaches that utilised their breadth of knowledge and expertise in image-based feature engineering, computational modelling, and machine learning. However, the model's performance was quite low, suggesting a weak association between structural MRI and residual intelligence scores at this age.

The winning submission was a combination of these and will be presented at a workshop as part of the MICCAI 2019 conference in China later this year.

Related links:

• ABCDNP Challenge: ‘Can you predict fluid intelligence from T1-weighed MRI?’
• Paper: ABCD Neurocognitive Prediction Challenge 2019: Predicting individual fluid intelligence scores from structural MRI using probabilistic segmentation and kernel ridge regression