Department of Computer Science · University of Oxford
Machine learning for the developing brain, built for ultrasound.
We develop machine learning methods that turn routine ultrasound into quantitative information about the developing brain.
Research themes
Four interconnected problems, from reconstructing 3D anatomy to understanding development after birth.
Computational ultrasound: geometry and physics
Geometric and physics-informed methods for ultrasound, from recovering 3D brain anatomy from freehand 2D scans to reducing artefacts such as acoustic shadows.
Multi-view geometry · physics-based rendering · inverse problems
Structure: brain atlases and segmentation
Building spatiotemporal atlases of the developing brain, and using them for registration, segmentation and shape modelling.
Segmentation · registration · statistical modelling
Robust everywhere: harmonisation and deployment
Building models that generalise across scanners, sites and clinical settings.
Domain shift · robustness · efficient and federated ML
From scan to outcome: brain growth and neurodevelopment
Modelling brain growth across gestation and linking prenatal imaging to neurodevelopment after birth.
Longitudinal modelling · biostatistics · clinical collaboration
Featured paper · Nature · 2023
Normative spatiotemporal fetal brain maturation with satisfactory development at 2 years
Using 3D ultrasound scans from the international INTERGROWTH-21st study, we built a normative atlas of how the fetal brain matures week by week, in pregnancies where the children went on to develop well at 2 years. The atlas gives clinicians and researchers a reference for what typical brain development looks like during pregnancy.
Latest news
All news →Joining the lab
We welcome enquiries from prospective DPhil and MSc students and postdocs. See how to apply.