There have been releases of high quality, large sample and open access studies of the neonatal and infant brain (exemplified in studies such as the developing and baby Human Connectome Projects). This focus is due to the importance of this early life period on later developmental outcome. Analysing these new datasets is extremely challenging, the brain develops extremely rapidly over the perinatal period and much more than any other time in postnatal life.
This development affects obvious factors like brain and structure volume, but also image intensity and local shape. This presents challenges in image processing and image statistics, potentially reducing sensitivity to real relationships between imaging and outcome. Standard neuroimaging pipelines, tailored for child or adult populations, do not easily work with these data. Developmental trajectories over short periods of life (weeks to months) are large highly non-linear so typical statistical approaches are imprecise or, in many cases, inappropriate.
The combination of these difficulties mean that analysis pipelines for the developing brain have been highly innovative. This challenge proposal will seek to draw and develop pipelines built at sites internationally into one place while addressing 2 benchmark questions in developmental neuroimaging:
1. Can we estimate the age at birth of preterm neonates scanned at later age.
2. Can we detect lesions or abnormalities in a way that is robust to the age of a neonate at time of scan.
For these challenges, we will use hold out data collected at the Evelina Newborn Imaging Centre. These images will be collected in a similar way to the the open access and freely available developing Human Connectome Project. Missing data is a reality in clinical imaging data and some contrasts will be missing for some of the holdout datasets, or will be collected in a different way (so 3D versus 2D acquisitions for MPRAGE images for example).