Diffusion MRI of the developing brain provides a unique means to study the emergence of the structural connectivity of the brain, and the impact of conditions such as preterm birth, autism, and epilepsy. Ultra-high field imaging on the KCL 7T system offers new opportunities for these types of investigations, with higher signal and stronger gradients providing access to higher diffusion weighting. This system also permits newer diffusion encoding strategies that provide sensitivity to different properties of the tissue.
However, most current analysis methods are based on simplified models of the tissue developed for the adult brain, which differs markedly from the developing brain. There is a clear need to develop advanced diffusion MRI analysis methods specifically for the developing brain, leveraging the increased signal and gradient strength available at ultra-high field.
The aims of this project are to:
(1) Adapt existing methods for the acquisition and reconstruction of high spatial and angular resolution multi-shell diffusion MRI data suitable for use in unsedated neonates on the 7T MRI scanner at St Thomas’ Hospital.
(2) Acquire pilot datasets of neonatal and infant diffusion MRI data at 7T to estimate connectivity in the developing brain in unprecedented detail.
(3) Develop models specifically tailored to the developing brain to provide information about the composition, orientation and other properties of the tissue in the developing brain.
(4) Investigate the use of cutting-edge AI-based segmentation and predictive models for these data using domain adaptation techniques to leverage large databases already available at lower field strengths.
In this project, the student will learn about the principles of diffusion MRI, neonatal brain development, advanced diffusion modelling algorithms, and deep learning approaches for image segmentation and predictive modelling.
Year 1 / MRes project: Set up a suitable diffusion MRI protocol on the 7T scanner. Acquire preliminary neonatal brain data to assess quality and feasibility.
Year 2: Refine acquisition protocol and process the data using established diffusion MRI pipelines to demonstrate the achievable data quality. Acquire pilot neonatal dataset to test and refine the image acquisition and subsequent analysis.
Year 3: Acquire data on a larger cohort. Develop advanced diffusion models specifically tailored at 7T neonatal diffusion data.
Year 4: Develop deep learning methods for the assessment of the emergence of connectivity in the developing brain. Enlarge training datasets with lower-field data using domain adaptation techniques.
