Project ID BE-MI2024_14

ThemeBE-MI

Co Supervisor 1A Faculty of Life Sciences & Medicine, School of Biomedical Engineering & Imaging Sciences, Department of Biomedical EngineeringWebsite

Co Supervisor 1B Institute of Psychiatry, Psychology & Neuroscience, School of Mental Health & Psychological Sciences, Department of Forensic and Neurodevelopmental ScienceWebsite

Additional Supervisor Dr Vanessa Kyriakopoulou

AI-driven simulations of brain development

Early fetal and infant neurodevelopment is a critical period, in which the human brain undergoes rapid growth and folding, underpinned by complex processes of cellular maturation and migration. Perturbations to this process, either through genetic or environmental influences can have profound implications. Better modelling of the causal impact of these processes is therefore vital, if we are to predict which infants are more likely to develop long-term neurodevelopmental difficulties.
Classical approaches to human brain analysis lack the precision needed to tailor predictive models to individual brains; however new AI techniques are opening a door towards sophisticated generative models of a range of neurological processes (Bass C NeurIPS 2020; Da Silva M. MLCN 2021; Fawaz A MIUA 2022)

This project will develop AI-driven mechanistic models of neurodevelopment, using MRI imaging data collected from individuals at multiple timepoints from fetus to 3-4 years of age. It will benefit from access to the Brain Imaging in Babies Study (BIBS), affiliated with AIMS2-trials, which recruits individuals at higher likelihood of later neurodevelopmental difficulties, such as autism.

Plan:
Years 1-2: Train in machine learning and cortical surface processing, for analysing human brain data; harmonise image processing across scan timepoints and build normative models of neurodevelopmental trajectories for different brain regions.

Year 2-3: Extend existing deep generative models of neonatal neurodevelopment (Fawaz A MIUA) with causal networks. This will support tailored simulation of neurodevelopmental trajectories from fetus to 4 years and offer mechanistic insight into which environmental and genetic risk factors most impact brain appearance. In this way it will be possible to suggest new biomarkers predictive of children’s neurodevelopmental outcomes at 4 years.

Year 4 – 6 months will be reserved for write up.

Representative Publications

1. ‘Glasser MF, Coalson TS, Robinson EC, Hacker CD, Harwell J, Yacoub E, Ugurbil K, Andersson J, Beckmann CF, Jenkinson M, Smith SM. A multi-modal parcellation of human cerebral cortex. Nature. 2016 Aug 11;536(7615):171-8.

2. Bass C, da Silva M, Sudre C, Tudosiu PD, Smith S, Robinson EC. ICAM: interpretable classification via disentangled representations and feature attribution mapping. Advances in Neural Information Processing Systems. 2020;33:7697-709.

3. Fawaz A, Williams LZ, Edwards AD, Robinson EC. A Deep Generative Model of Neonatal Cortical Surface Development. In Medical Image Understanding and Analysis: 26th Annual Conference, MIUA 2022, Cambridge, UK, July 27–29, 2022, Proceedings 2022 Jul 25 (pp. 469-481). Cham: Springer International Publishing.

1. Ciarrusta J, Dimitrova R, Batalle D, O’Muircheartaigh J, Cordero-Grande L, Price A, Hughes E, Kangas J, Perry E, Javed A, Demilew J. Emerging functional connectivity differences in newborn infants vulnerable to autism spectrum disorders. Translational psychiatry. 2020 May 6;10(1):131.

2. Andrews DS, Marquand A, Ecker C, McAlonan G. Using pattern classification to identify brain imaging markers in autism spectrum disorder. Biomarkers in Psychiatry. 2018:413-36.

3. Kevin KY, Cheung C, Chua SE, McAlonan GM. Can Asperger syndrome be distinguished from autism? An anatomic likelihood meta-analysis of MRI studies. Journal of Psychiatry and Neuroscience. 2011 Nov 1;36(6):412-21.