Using hallucinations to understand top-down processes in human vision.

The overarching aim of the project will be to empirically assess the spatial and temporal
dynamics of top-down signals giving rise to visual illusions, with a specific focus on
hallucinations.
The student will have the opportunity to explore these processes in healthy controls
experiencing sensory deprivation, for example using fine-grained controlled visual
stimulation to induce hallucinations, or whilst experiencing chemically induced hallucinations.
For the latter, the student will have access to and contribute to the collection of a unique
dataset of people taking part in a DMT study, undergoing a single-dose acute challenge with
intravenous DMT fumarate in the MRI scanner, versus a healthy control group (N=20). There
is also an exciting opportunity to investigate patients with severe visual impairments
experiencing Charles Bonnet Syndrome, who can be recruited via Dr. Dekker's connections
with Moorfields Eye Hospital and Charity.
Years 1-2. Write literature review on internally-generated vision, and plan, program, and pilot
fMRI/behavioural experiment on the dynamics of visual illusions and imagery in healthy
controls and/or patients with Charles Bonnet Syndrome. Support neuroimage collection for
DMT study (supervised by both Skipper/Dekker). In these studies, we acquire fMRI both on
and off hallucination and under matching circumstances. On two separate days prior to and
subsequent to this scan session, participants will watch movies and images during fMRI to
assess neural processing during externally driven vision.
Years 2-3. Complete all data collection. Analyse data from visual hallucination datasets by
developing connective field modelling [5] and more complex encoding modeling and machine
learning approaches [6,7] to compare visual processing dynamics during hallucinations,
movie watching, and rest. We will also aim to use labelled networks in machine learning
approaches (like LTSM) to determine which network dynamics best predict hallucination
content and vividness.
The feasibility of the project is demonstrated by the successful collection of N=86
participants watching full length movies during fMRI during the same time frame in a prior
LIDo PhD project by supervisor Skipper, which established graph-theoretic and machine
learning based approaches (for further details, see [8]
https://www.nature.com/articles/s41597-020-00680-2). Moreover, the DMT project has
already received ethics approval (ID number 17715/001), granted by the UCL Research Ethics
Committee, from the Office of the Vice Provost Research. NHS ethics will need to be sought
for research with vision patients. This, and recruitment can take long, but is factored into the
timeline and mitigated by access to other interesting participant groups and approaches.