Student: Russell Macleod
The aim of this study is to be able to provide individualised markers of brain tissue abnormalities in complex disorders of neurodevelopment such as childhood epilepsy. This will not only provide treatment targets, it will guide individually tailored interventions (i.e. precision medicine).
Detecting these markers using neuroimaging methods is hampered by the rapid developmental changes in the brain throughout infancy and beyond. In practice, this means that sensitivity to neuroanatomical changes in childhood disorders can vary depending on what age the child is. This heterogeneity can either mask true changes in clinical studies and trials or bias studies towards very circumscribed age-ranges.
This project will develop and apply machine learning techniques to model typical brain development using large MRI datasets of >2000 infants, toddlers and children, and new data collected in children with epilepsy. The resulting model will be used to maximise detection of abnormal tissue in individual children.
Epilepsy and autism spectrum disorders are two common serious disorders of neurodevelopment. They can co-occur and are associated with structural and functional changes in the brain. Both disorders are complex, occur over neurodevelopment and, reflecting this, brain changes associated with these disorders can be highly variable from person to person. For infants and children especially, our ability to detect these atypical changes in the brain using neuroimaging are hampered by the ongoing typical neurodevelopment. This effectively means that the age of a child can determine the sensitivity of neuroimaging to subtle or even gross pathology.
Aims and Importance:
This project will aim to develop a normative model and atlas of the developing brain using magnetic resonance imaging (MRI) and statistical modelling. Using advanced quantitative MRI relaxometry, the project will expand this atlas to be adaptable to different MRI sites and scanners, ensuring that the resulting work can be used translated beyond specialist research sites. Finally, it will use this atlas to investigate tissue abnormalities in individual children with epilepsy and autism, two related neurodevelopmental disorders, with the intention of providing single subject inference of brain tissue contrast and volume.
Planned research methods and training provided:
The candidate will build statistical models of image intensity changes occurring over infancy and childhood, initially using pre-existing structural MRI data. They will receive training in image analysis and registration and neurobiological aspects of brain development. The candidate will also develop pipelines, inspired by work in statistical genetics, for combining datasets from multiple different sites and scanners.
They will be trained in supervised and unsupervised machine learning techniques applied to medical imaging. The supervision team have strong experience in developmental imaging, machine learning techniques as well as the application of these techniques to neonatal and childhood neurology. This project will also be highly collaborative, involving interactions with clinicians, physicists, image scientists and statisticians.
Objectives / project plan:
Year 1: Training in statistical packages (especially in python) and image registration techniques. Begin collecting MRI data (T1 and T2 relaxometry) in children (both patients and healthy volunteers).
Year 2: Continuation of data collection. Build computational models of brain development from pre-existing data.
Year 3: Completion of data analysis. Optimise tissue pathology detection using quantitative MRI.
Year 4: Finalize image analysis and complete write-up.