Cookies on this website
We use cookies to ensure that we give you the best experience on our website. If you click 'Continue' we'll assume that you are happy to receive all cookies and you won't see this message again. Click 'Find out more' for information on how to change your cookie settings.
Skip to main content

EXTERNAL SUPERVISOR

Dr. Habib Ganjgahi, Department of Statistics, University of Oxford

Background

The last decades of neuroimaging research have brought immense insight into our understating of human brain function and structure. However, these findings are limited by the use of small and unrepresentative samples. The UK Biobank (UKB) is a landmark study based on ½ million volunteers, 100,000 of which will undergo magnetic resonance imaging (MRI) scanning. The data collected will address the sample size concern in imaging studies and advance our understanding the biological basis of neurological and psychiatric illnesses. The UKB projects collects multiple types of MRI data of the brain, genetic data and 1000’s of behavioural, environmental, and health variables on each subject, giving enormous opportunity to understand the individual variation in the function and structure of human brain.

There are biological and non-biological factors that explain variability in neuroimaging studies. It has been shown that correcting for age, sex, age2 and their interaction dramatically affects association analysis between imaging and non-imaging phenotypes in the UKBB. Moreover, non-biological factors like head motion during scan, head and table position and slow drift in imaging phenotypes distribution over time due to scanner degradation can explain additional variability. While a traditional regression approach can be a starting place for controlling for these variables, this comes with linearity and additivity assumptions as well as model selection challenges.

The purpose of this study is to evaluate Gaussian process (GP) techniques for modelling nuisance variation in large scale neuroimaging studies. GPs are a powerful tool to capture non-linear relationships that can be used as a prior in Bayesian analysis. However, scalability is a major concern in GP.  The aim of this project is to exploit the recent advances in scalable gaussian processes and Bayesian mixed effect model to deconfound and ultimately harmonise imaging phenotypes in the UKBB. 

RESEARCH EXPERIENCE, RESEARCH METHODS AND TRAINING

The successful completion of the project will likely produce at least one publication and conference presentations.  At least an undergraduate degree in mathematics, statistics or similar technical subject and good writing skills are required, and will be supplemented by participation in the Academic for PhD Training in Statistics (APTS, http://apts.ac.uk) candidate requires to participate in APTS and modules related to the project. 

PROSPECTIVE CANDIDATE

Students with training background in mathematics or statistics; experience with programming. 

Supervisor

  • Thomas Nichols
    Thomas Nichols

    Professor of Neuroimaging Statistics, Nuffield Department of Population Health