Characterizing Spatiotemporal Dynamics of Fetal Brain Development from Ultrasound Images

To join this seminar online, please see

Ultrasound (US) imaging is one of the first steps in a continuum of pregnancy care, yet due to the difficulty of interpreting the images, it is not routinely used as a neuroimaging modality. During the fetal period, the brain undergoes dramatic structural changes occurring in the scale of weeks. The resolution of modern US machines enables us to observe such changes in fetuses from as early as 14 gestational weeks. Capitalising on recent breakthroughs in deep learning, our group develops bespoke methods to automatically process neurosonographic images and track the spatio and temporal patterns of intra-uterine brain development. We have built the first US-based atlas of the fetal brain from the global INTERGROWTH-21st dataset, which reveals previously undescribed structural details and emergence of asymmetries between cerebral hemispheres. I will be discussing our data processing pipeline and our approach to characterising healthy, in-utero developmental patterns. We envision that this atlas will serve as a population reference against which individuals can be compared from the early stages of pregnancy, and hence enable detection of developmental deviations in routine clinical care.