Identifying Causal Interactions In Functional Mri Data Using Statistical Time Series Analysis

Our understanding of the brain has been revolutionized by functional neuroimaging, the ability to look at brain activity as a person is doing a task, or even while they do nothing at all. Traditionally, neuroscientists have used neuroimaging to localize different functions to different parts of the brain. However, brain functions depend on dynamical networks spanning many different brain regions. Identifying these networks, and especially networks that show causal interactions among their elements, is a major current challenge. This Ph.D. project will address this challenge for functional MRI (fMRI), the most popular neuroimaging method. fMRI measures time-varying changes in metabolic signatures of neural activity. To identify causal networks, we adopt the framework of Granger causality analysis (GCA) which assumes that causes both precede and help predict their effects. GCA applied to fMRI faces several challenges arising because the fMRI signal is an indirect and incompletely understood reflection of underlying neural activity, is sluggish, delayed, and is sampled only once every 2-3 seconds. The project will address these limitations by novel combinations of theory, modelling and experiment. A first objective will be to adapt recent theoretical findings showing invariance of GCA under filtering to the case of fMRI. A second objective will be to characterize the behavior of GCA on fMRI data via detailed computational models connecting neural activity to simulated fMRI responses. These models will be built by connecting existing large-scale spiking neuron simulations with forward models of hemodynamic responses. Third and finally, the resulting methods will be benchmarked on fMRI data obtained specifically for this purpose.

The successful candidate will benefit from being part of the Sackler Centre for Consciousness Science, a world-leading research group in computational neuroscience, functional analysis of neuroimaging data, and consciousness research. The student will enjoy collaborative input from the Department of Informatics, the Brighton and Sussex Medical School, and the School of Psychology; will work within a thriving multidisciplinary group integrating many areas of neuroscience, and will have full access to state-of-the-art computational and neuroimaging facilities. Training in fMRI analysis and statistical methods will be provided.

Applications should hold, or expect to obtain, a minimum upper-second honours degree (or equivalent) in a quantitative science discipline. Previous experience in neuroimaging and/or time series analysis is desirable but not required.