The vast majority of acquired amputees report persistent and vivid sensations of their missing limb, with most amputees reporting that these sensations are painful. Neuroimaging work has demonstrated the preserved representation of the phantom limb in the sensorimotor cortices, and the extent of preservation (i.e. activation) has been linked to the incidence of phantom limb pain (PLP), with movement of the phantom hand engaging activity in typical pain processing areas. This body of literature suggests that an interplay of preserved representation in phantom hand motor control and pain processing may be involved in the generation of PLP. However, it is still uncertain how these factors relate to one another and whether phantom pain lies at one end of a phantom sensation spectrum. With this project we aim to develop new paradigms in order to probe the mechanisms underlying the generation of phantom sensations and PLP both within and beyond the sensorimotor system. This will involve a range of techniques from functional MRI, Transcranial Magnetic Stimulation, high-density EMG, OPM-MEG and computational modelling (drift diffusion modelling, reinforcement learning). We are specifically interested to test (using pre-registered reports) the plausibility of using a predictive coding framework to account for inter-individual differences in PLP.

Using a cohort of healthy adults from 18-88 years of age in the CamCAN project (https://www.cam-can.org/), we have shown that mid-life activities provide resilience of cognitive abilities to age-related atrophy of gray-matter in old age (from structural MRI [1]). This project will investigate the neural and vascular correlates of this “cognitive reserve” using multimodal, longitudinal neuroimaging data, including fMRI and MEG, as well as lifestyle and health data, from the CamCAN cohort and other European cohorts within the LifeBrain consortium (https://www.lifebrain.uio.no/). This project would benefit from good statistical (eg multivariate) skills and basic knowledge of neuroimaging, though both of these can also be taught.

In recent years, there has been increased interest in augmentative technologies that enhance the physical and cognitive abilities of the human body. These innovative devices introduce various theoretical and practical neuroscience challenges: What resources can the brain employ to control a body part that has never been there before? In this project, we work with the Third Thumb (Dani Clode Design) – a robotic finger, designed to allow the user to single-handedly perform typically bimanual tasks. This technology provides innovative solutions for increasing the functionality of disabled individuals in daily life, including stroke patients, children with developmental hand malformations and even people sustaining temporary arm injury. We will recruit individuals from these patient groups and train them to use the Third Thumb, using a combination of at-home and lab-based training programmes. To facilitate successful Thumb skill learning and generalisation, we will need to consider key cognitive principles relevant to strategy, explicit knowledge, adaptation, reinforcement learning, abstraction of motor planning, motivation, multitasking, interference and agency – all necessary for successful implementation, and mastery, of motor augmentation. We will use a range of behavioural, physiological, kinematic and neuroimaging techniques to better assess the neural basis of successful integration of the augmentation device into the human body representation.

More than a third of the autistic population does not develop functional spoken language. Yet, mounting case studies and neuroimaging evidence suggest that their language comprehension and semantic processing may be preserved. This raises the question of which cognitive processes are disrupted in the language comprehension-to-production pathways. Using and machine learning approaches on neuroimaging data (e.g., EEG, MEG), we propose to examine the content, timing and exchange of information in the brain while language is processed and transformed into motor commands. This will be crucial both to inform neurobiological theories of language and motor control, and for shaping future interventions in the minimally-verbal autistic population.

Early detection of Alzheimer’s disease (AD) is essential for developing effective treatments. Neuroimaging techniques like Magnetic Resonance Imaging (MRI) have the potential to detect brain changes before symptoms emerge [1]. Structural MRI can detect atrophy related to AD, but functional changes related to loss of synapses and synaptic plasticity occur even earlier. Our work tests the potential of Magnetoencephalography (MEG) to detect progressive differences in brain activity in people with early stages of dementia, such as Mild Cognitive Impairment (MCI), or even healthy people at-risk of AD owing to lifestyle or genetic factors. We use machine learning to extract features from the rich spatiotemporal dynamics of resting-state MEG, and the fitting of neurophysiological models (such as Dynamic Causal Modelling) to task-evoked responses [2]. A PhD within this area would involve developing and validating advanced MEG analyses on a number of large, pre-existing datasets, such as BioFIND, NTAD and CamCAN.

When people prepare for a forthcoming task, they spontaneously ‘clear their mind’ of aspects of the previous task and assemble what is needed for the upcoming task (e.g. Farooqui and Manly, 2019). Such transitions also play a role in memory. For example, event boundaries are associated with spontaneous increases of activity in the hippocampus (Ben-Yakov and Henson, 2018) and may help account for why surprising events (that abruptly signal a new, unexpected episode) are so memorable (Ben-Yakov, Smith and Henson, in press). The reports of some people with acquired brain injuries (ABI) are consistent with a greater difficulty in segmenting experience into such discrete events. Behaviours like perseveration (in which actions relevant to a previous goal are produced without obvious purpose to a current goal) are also suggestive of greater permeability in episode boundaries. This project will examine relationships between cognitive preparation for upcoming tasks, segmentation (e.g. of films into discrete events), and its effect on memory in ABI. It will also consider people’s (with and without ABI) patterns of segmenting their own experience. Of particular interest is whether greater attention to task and context transitions, or training in more volitional segmentation, could lead to improved memory performance.

Our previous work (Farooqui and Manly, 2019) has demonstrated a set of interesting apparently non-volitional phenomena when people perceive themselves as beginning a new task (even if the task is, in fact, a continuation of the same activity). These include clearing out representations/rules etc. from the previous task and a time cost in preparing for the new episode that pays off in terms of

enhanced cognitive control. There are many important basic questions including whether these effects generalise to other tasks, the length of episode over which such planning can operate, neural correlates/dynamics of these preparatory acts and whether, for example, a perceived change in context similarly triggers the creation of a new ‘task episode’. Many mental health conditions are characterised by intrusive thoughts unrelated to (and unhelpful to) a current goal and it is of interest to investigate whether the creation of discrete task boundaries is useful in reducing this interference.  Adaptive features of segmenting activity into discrete ‘task episodes’ raise interesting questions about whether these processes are compromised by brain injury and whether this provides useful additional accounts of features such a distractibility, goal neglect and perseveration as well as offering potential new targets for training.

Listeners with intact hearing achieve a near-optimal combination of speed and accuracy in identifying spoken words. Words can often be recognised before their offset, time-locked to the earliest sound (/z/) that distinguishes a target word (lizard) from similar sounding neighbours (listen). However, for degraded speech or hearing-impaired individuals, word recognition becomes less efficient even if recognition accuracy remains high. For example, speech-driven eye-movements show delayed identification of words like lizard due to consideration of rhyming words (wizard) [1]. The proposed project will devise behavioural methods to measure the speed and efficiency of spoken word recognition in individuals with intact or impaired hearing, including cochlear implant users. EEG and MEG measures of brain activity during spoken word recognition can be used to constrain neural and computational theories of speech identification [2]. Assessing the efficiency of spoken word recognition allows us to quantify processes that are critical for everyday comprehension and that minimise listening effort. This provides an initial step towards quantifying and improving real-world listening outcomes for hearing-impaired individuals.

Difficulty/slowness in thinking of the right word is a common and frustrating symptom of stroke. Whilst disruption to language systems is clearly a major component in such anomia, there has been debate about the contribution of more general cognitive control processes (sometimes called ‘executive functions’) in contributing to word retrieval (e.g. Schumacher et al., 2022). There are good reasons why this might be the case. Firstly, word retrieval in healthy people can be slowed by concurrent executive demands (e.g. dual task situations). Secondly, executive functions are particularly important for developing new strategies to meet novel challenges. These can be seen as forming a kind of ‘cognitive scaffolding’, allowing repeated practice of a task until the coordination of necessary processes becomes incrementally more automatic. To date, investigations of executive function – post-stroke anomia links have largely been correlational (the co-incidence of executive and language impairments), or have looked at disruptive influences of dual task demands on fluency (e.g. Gehman and Faroqi-Shah, 2019). The current project will examine potential facilitation of word finding via interventions that may increase available executive capacity (e.g. Farooqui and Manly, 2019) and whether such techniques have potential application in enhancing rehabilitation efficacy. 

There is widespread agreement that Bayesian inference – combining sensory signals with prior knowledge or predictions – is critical for the perception of everyday sounds such as speech. Evidence from fMRI and MEG studies support an account in which a key computation in superior temporal gyrus (STG) involves comparing prior expectations (predictions), with sensory signals (speech sounds) and using the resulting prediction errors to update higher-level representations and perceptual outcomes [1]. However, the cortical circuits that combine predictions and speech sounds in the STG remain under-specified. We thus have an incomplete understanding of how these processes support successful speech perception, or how they are challenged when we misperceive degraded sounds, or when individuals prone to psychosis experience speech sounds that are not physically present (hallucinations). This project will combine high-resolution, ultra-high field 7T fMRI with multi-voxel pattern analysis methods to specify the laminar organisation of the neural representations of predictions, speech sounds and prediction errors in the STG. In the process we can more fully specify the cortical operations that produce our perceptual experience of speech.

Functional imaging and MEG studies show additional activity in inferior frontal and inferior temporal regions during comprehension of ambiguous words in spoken sentences (e.g. “The steak/stake was rare just as the customer had requested”) compared to matched low ambiguity sentences [1]. Although fronto-temporal activity is consistently associated with ambiguity resolution, this does not show that these regions play a causal or specific role in the interpretation of ambiguous words. Causal evidence can come from studies that assess whether brain lesions or stimulation of inferior frontal and inferior temporal regions causes impaired comprehension of ambiguous sentences. This project will devise behavioural measures of the comprehension of ambiguous words in sentences that can be administered to individuals with fronto-temporal lesions (from the CCNRP cohort) or combined with transcranial magnetic stimulation. These studies can thereby assess the causal role of inferior frontal and inferior temporal regions to comprehension of ambiguous sentences. By considering whether these effects are unique to ambiguity resolution, apply to a range of linguistic or semantic challenges, or overlap with domain general executive functions (e.g. working memory, selection) we can determine whether fronto-temporal brain regions make a specific, causal contribution to ambiguity resolution. Combining brain imaging with lesions or stimulation allows assessment of resilience and plasticity following neural perturbation with implications for rehabilitation of sentence comprehension.

Cochlear implants (CIs) restore hearing to deaf people by electrically stimulating the auditory nerve. Although many CI listeners understand speech well, the variation in outcomes is large and some patients struggle even in quiet backgrounds. Furthermore, new methods of programming CIs also vary in their effectiveness, with some patients showing a substantial benefit while others show no benefit or even perform worse with the new method.  Previous attempts to understand the factors underlying good and bad performance have correlated speech scores with the outcomes of various sensory and cognitive tests and with demographic factors; they have been largely unsuccessful probably because the different tests correlate with each other. The proposed project will adopt a different approach which is to separately manipulate sensory and cognitive factors, for example by blurring speech sounds and imposing a secondary non-auditory task. We then study how these factors affect speech perception on their own and on how and whether they interact. The experiments will initially be performed with normal-hearing (NH) people listening to simulations of CI hearing and then applied directly to CI listeners. It builds on our previous work showing how cognitive factors such as attention can sharpen sensory representations, and help us understand why methods for improving CI hearing, including those developed in our lab, help some patients more than others. Eventually we hope to apply the findings to provide patient-specific solutions to improving speech perception by CI users.

A promising approach in management of anxiety disorders is self-distancing .  In this procedure, the person imagines a scene that induces anxiety, then increasingly “pulls back” to imagine themselves further and further away.  This kind of change in an imagined spatial context may depend on the brain’s default mode network (DMN), and more widely, we have evidence that the DMN is involved in large changes of mental focus. Based on supportive pilot data, we should like to test the role of broad DMN involvement in escape from a mental focus on threatening mental contents in the general population, in individuals with mental health problems, and in people with acquired brain injuries that may disrupt relevant underlying processes.

This project will examine (mal)adaptive social media use in clinical populations or those at risk of a clinical diagnosis, an area of substantial concern where pre-existing theoretical approaches can substantially inform discovery science and translation (RCPsych, 2019).

1) Transdiagnostic approaches: A PhD project could study the influence of social media use in a) predicting adolescents receiving a mental health diagnosis, b) impacting those living with a mental health diagnosis, or c) determining recovery. Of particular interest is how we can approach these questions from a transdiagnostic perspective.

2) Mechanisms: A PhD project could focus on detailed investigations using theoretical and cognitive approaches to understand the mechanisms behind how social media exacerbates mental health problems, especially affective disorders. Experimental or neuroimaging studies could be used to test whether features of social media exacerbate maladaptive cognitive mechanisms prominent in affective disorders, e.g., the influence of prediction errors in social settings and rumination.

3) Clinical Contexts: A project could focus on how social media use is assessed and examined in a clinical context, and how such processes could be improved and supported. A lot of previous work in this research area has relied on examining population averages, however we have made very little progress in helping individuals. A PhD project could focus on how to integrate assessment and adjustment of social media use into clinical treatment or how to develop methodologies to help adolescents on an individual basis.

Acquired brain injuries (ABI) are a leading cause of long-term disability, with cognitive, mood and social difficulties being common impairments. Traumatic brain injuries (TBI; e.g. from car and sports accidents) are among the most common causes of ABI and particularly affect the young (for example, each year approximately 50,000 US adolescents are hospitalised with TBI; Asemota et al., 2013). The negative impact of such injuries on the development of peer relationships and for mental health in general is well documented. Relatively little is known, however, about the use, risks and benefits of the digital world in people with ABI. Important questions include whether cognitive impairments form barriers to social media engagement, whether individuals isolated from local peer support are able to find this online, and potential risks to self-esteem and mental health. Other questions include the use of digital devices to support cognitive function, to reinforce rehabilitation goals and to track mood and activity with reduced requirements for accurate memory and self-report.

Unilateral spatial neglect, in which individuals have difficulty in attending to/respond to one side of space, is a very common consequence of stroke. This high frequency, in combination with the very wide variety of brain structures where damage has apparently led to neglect, suggests that it may be better thought of as a ‘state’ into which the brain is prone to fall rather than an impairment to an encapsulated module. In line with this, healthy adults appear to show neglect-like phenomena in states of drowsy sleep onset (Bareham et al., 2014) and after sleep deprivation (Manly et al., 2005), and the degree of bias shown by patients with neglect can be significantly modulated by changes in alertness levels (e.g. Robertson et al., 1997; George et al., 2008). Reductions in neglect have also been reported in the context of pleasant stimuli (Soto et al., 2009) and reward (Malhotra et al.,2012), suggesting that current mood also plays a role. This is potentially important in understanding post-stroke recovery patterns because low arousal, reduced activity and mood disorder are common and complex interactive factors following acquired brain injury (low alertness <-> low activity levels <-> low reinforcement <-> low mood). On this basis, we might predict that other clinical groups, primarily defined by low mood, may also be vulnerable to showing spatial biases. A first step in this project would be to follow up on preliminary data, investigating performance on sensitive spatial measures in individuals with low mood. Further questions concern whether mood and alertness components are separable, underlying neural dynamics, and the efficacy of interventions.

Acquired brain injury (ABIs), such as from stroke or traumatic brain injuries, can cause dramatic changes in awareness of body parts and sensations. ABIs are also associated with an almost 5-fold increased risk of mood disorder compared to people from the general population. To date, we know little about links between bodily sensation and mood following ABI, for example, whether distorted/changed interoceptive signals influence affective experiences and how mood changes may modulate sensation and awareness. Given our current lack of awareness, a useful project would include interviews with ABI survivors about changes in sensation and awareness and the adaptation of existing measures for survey within this group. This should give rise to tractable hypotheses for experimental manipulation, for example, the effect of mood enhancing contexts on bodily awareness.

Can we build simple mathematical models that capture longitudinal changes in the developing brain? The human brain has a characteristic organisation, including hallmarks like heavily connected hub regions. How and when does this organisation emerge, and how does it change as children develop? We have collected multiple longitudinal datasets here at the CBU, including the CALM cohort (https://calm.mrc-cbu.cam.ac.uk/), and now we want to see if we can build simply mathematical models ⁽¹⁾ that capture how brain networks change across two time points spaced 2-5 year apart.  By producing models that recapitulate the real changes that appear in observed brains, our aim is to understand the underlying principles that shape development and give rise to the diversity that we see across individuals, including those with neurodevelopmental disorders.

Genome-wide diagnostic testing is now widely implemented in clinical practice to identify the underlying cause of neurodevelopmental disorders.  Diagnostic yields are increasing, meaning that the number of individuals and families receiving a genetic diagnosis will climb over the next few years.  The challenge now is to make genetic diagnosis more useful for affected individuals and their families.  To achieve this, we need a much better understanding of the links between genetic cause and an individual’s symptoms and impairments.  One approach is to group genetic diagnoses into functional networks converging on molecular and cellular processes such as chromatin regulation, ion channels / excitability and synaptic signalling.  Functional networks have distinct cellular, spatial and temporal expression patterns, suggesting that they may impact on different neurodevelopmental mechanisms.  We want to know whether, how and why functional networks differentially influence cognitive development and mental health.  Ultimately, the goal is to provide evidence-based prognostic information for each person after genetic diagnosis, and translate this knowledge into personalised mechanism-informed therapies which can improve specific symptom dimensions of value to each person.  This project will involve big data (for example, the 100,000 Genomes Project, IMAGINE-ID, Simons Searchlight and SPARK) and computational analyses integrating genomic, cellular and phenotypic datasets.

Chromatinopathies are neurodevelopmental disorders arising from rare, pathogenic variants in regulators of DNA structure, impacting on gene transcription.  Collectively, chromatinopathies are a relatively common cause of intellectual disabilities, autism spectrum disorders and other associated cognitive and behavioural characteristics. In our previous functional networks study of autism spectrum characteristics, we found that individuals with chromatinopathies had elevated symptoms of behavioural inflexibility.  We have now repeated this analysis in a larger independent cohort, and found again that the chromatinopathy group have a distinctive dimensional profile involving social relating / communication and sensory / motor flexibility.  This project will build on these findings by investigating the cognitive and neural systems characteristics of chromatinopathies.  Methods may include cognitive testing (using our recently published remote, dynamic cognitive testing app) and neuroimaging (using MRI and MEG).  We will apply advanced analysis methods such as generative modelling and HMM to investigate dynamic aspects of brain structure and function.

Focused Ultrasound Stimulation (FUS) is a non-invasive neurostimulation technique that has the potential to reversibly and safely modulate neural activity in regions of the human brain [1] that are difficult to target precisely with more traditional techniques like Transcranial Magnetic Stimulation (TMS). These include regions such as the hippocampus that are deep in the brain and so beyond the limited reach of TMS, and regions such as the anterior temporal lobes that are difficult to target with TMS without causing peripheral nerve stimulation. FUS has the potential to transform basic science by enabling causal inference about the functional properties of specific brain regions, as well as offer potential new treatments for dementia, stroke and mental health. This PhD project will combine FUS with behavioural and fMRI measurements, and could be taken in several scientific directions:

(1) Testing whether FUS can modulate activity in hippocampus during episodic memory tasks and help dementia (primarily supervised by Rik Henson, in collaboration with Drs Nord and Halai);

(2) Testing whether FUS could modulate anterior temporal lobe activity during semantic memory tasks and potential use in conjunction with stroke rehabilitation targeting perilesional tissue or the contralateral hemisphere (primarily supervised by Ajay Halai, in collaboration with Prof Henson and Dr Nord);

(3) Testing whether FUS modulates activity in subregions of the insula, measuring subsequent or concurrent effects on interoception in healthy patients and those with mental health conditions (primarily supervised by Camilla Nord, in collaboration with Prof Henson and Dr Halai).

[1] Tyler WJ, Lani SW, Hwang GM. Ultrasonic modulation of neural circuit activity. Curr Opin Neurobiol. 2018 Jun;50:222-231. doi: 10.1016/j.conb.2018.04.011. Epub 2018 Apr 16. PMID: 29674264.

Brain connectivity between brain regions is commonly computed based on univariate signals, i.e. activity patterns are collapsed across voxels per region. This neglects possible multidimensional relationships between these patterns. Furthermore, it does not provide information about the transformations of activity patterns between brain areas and across time. In this project, you would evaluate and advance recently proposed methods to estimate multidimensional brain connectivity using pattern transformations. Methodological challenges are the application of these methods to event-related designs and the limited spatial resolution in the case of EEG and MEG. These methods could be applied to timely questions in cognitive neuroscience using newly acquired datasets or existing open large-scale datasets.

It is well known that the spatial resolution of EEG/MEG source estimation is limited. This affects all applications that are based on source estimates. Most previous evaluations of source estimation methods have focussed on univariate activation, i.e. homogenous activation within brain regions. In this project, you would evaluate distributed source methods for applications of multivariate and multidimensional methods (e.g. pattern classification and multidimensional connectivity). For linear methods, this could be achieved by extending the concept of the resolution matrix (and point-spread and cross-talk functions) to activity patterns. This will require a range of systematic simulation studies, but there will be ample scope to test the methods on new or existing data sets.

Reading is a complex behaviour that requires the control of eye movements to optimally sample information from sequences of words. Previous neuroscientific studies have largely ignored the behavioural aspects of reading, and most previous evidence stems from simplistic word-by-word or single-word paradigms. In this study, you would investigate the dynamic interaction of psycholinguistic single-word variables (e.g. word frequency, concreteness) with contextual factors (e.g. predictability, plausibility) using co-registered eye-tracking and EEG/MEG. This would reveal the dynamic connectivity of brain areas that lead from visual and semantic processing and context integration to motor control, and may allow the classification of reading styles in individual readers.

Motion remains a prevalent issue within MR imaging potentially leading to significant artefacts and unintended smoothing. This is of particular importance when working with patients with movement disorders or dementia, or with ultra-high field imaging with submillimetre resolution. The standard quality control measures used in some MRI studies to reject or correct data based on movement parameters introduce a significant risk of bias. Motion correction at the time of data acquisition would be a better solution. This project will focus on the development of methods for prospective motion correction using a camera set up inside the MRI which tracks head movements in real time. The optimised methods will be applied to clinical translational studies in order to achieve smaller differences in motion patterns between groups; increase the signal to noise for differential diagnosis; and distinguish true brain differences from artefacts resulting from motion pattern.

Layer-specific functional MRI (fMRI) studies aim to address questions on directional flow of information between brain areas located at different cortical depths. In recent years, the advances in ultra-high field MRI have enabled the study of layer-dependent activation in the human cortex non-invasively. However, several methodological challenges remain when it comes to data analysis. For example, studies using BOLD (Blood Oxygen Level Dependent) contrast are affected by a signal bias towards the superficial layers due to the presence of large draining veins but there is no consensus on how to correct for such effects. There are also several options for artefact correction, as well as different methods for segmentation of cortical layers, most of which do not take into account the correspondence between cortical depth and cytoarchitectonical cortical layers. The aim of this project is to explore the different options for data analysis in fMRI layer-dependent studies and develop optimised analysis pipelines which can be applied to 7T MRI data analysis with minimal manual input from the user.

The locus coeruleus in the brainstem provides all of the brain’s noradrenaline, which is essential for attention, arousal, motor control, and cognition. This small nucleus, approximately 20x1x1mm in humans, is one of the earliest parts of the brain to be damaged by Alzheimer’s disease and Parkinson’s disease.  Because of the iron content of the neuromelanin made in noradrenergic cells, the locus coeruleus can be identified and quantified by magnetisation transfer sequences in MRI.  We have led major studies of imaging the locus coeruleus using ultrahigh field MRI (7T), linked to drug studies of people with movement disorders and dementia.  However, the signal to noise and resolution of such sequences has been challenging at 3T.  Sensitive scanning of the locus coeruleus at 3T would have many advantages, and a major impact in neuroscience and its clinical applications, because of the greater availability and safety of 3T MRI.  This project will therefore optimise MRI for locus coeruleus imaging at 3T, in health and clinical cohorts.

The human brain has a remarkable capacity for reorganisation in early life. Deprivation such as congenital deafness or hand loss is the most prominent model we have for understanding this reorganisation. Yet paradoxically, existing evidence comes almost exclusively from studies with adults, which cannot retrospectively tell us how the reorganisation took place. A key possibility is that altered behaviour during childhood, which children develop in order to cope with their disability, is a driver for brain reorganisation. Therefore, to understand the extent to which behaviour shapes functional brain organisation, we will study the brain as new behaviour is being developed. We will measure functional adaptations that children born with one hand develop in order to compensate for the lost functionality of their missing hand, and how this adapted behaviour might impact cognitive development. We will study adults who have lost a hand at different stages of development, from infancy to childhood, to determine the extent to which experience of having a hand stabilises brain organisation. We will use sign language in novices and expert signers to determine how motor experiences are transformed by linguistic experience. To ascertain the bidirectional relationship between reorganisation observed in the sensorimotor cortex and these compensatory behaviours, we will fMRI (both standard and ultra-high field), MEG and behaviour, in combination with advanced multivariate pattern analyses.

Longitudinal cohort MRI studies are becoming increasingly popular. In these studies, the same individuals are scanned multiple times over a number of years, or even decades, allowing us to study how their brains change as they develop (e.g. CALM https://calm.mrc-cbu.cam.ac.uk/) or age (e.g. CamCAN https://www.cam-can.org/). Therefore, such studies can generate extremely rich multi-modal MRI datasets, but there can be significant challenges as well. In particular, the longer the time between multiple scanning time points, the more likely it is that the scanner(s) involved will have been through a number of software and hardware changes, some planned and some unplanned. Even when imaging parameters are matched as closely as possible before and after any scanner changes, significant software or hardware changes will undoubtedly result in differences in data quality. Because these differences are inherently correlated with the changes we are hoping to measure – e.g., the effects of development/ageing in brain structure and function – robust data harmonisation pipelines need to be developed before combining data from multiple scanning time-points. Otherwise we run the risk of over- or under-estimating the magnitude of any effects we find. A number of data harmonisation algorithms for MRI data already exist in the literature (see for example Fortin et al., 2017, and Tax et al., 2019), so the first step for this project will be the evaluation of the different algorithms available using a ‘travelling heads dataset’ (i.e., a dataset where the same people were scanned across two different scanners). Once suitable harmonisation procedures have been identified, either by applying existing algorithms or developing new ones, these methods will then be used to combine multiple scanning time points for existing datasets for the study of brain development and/or ageing.

Brain stimulation using transcranial Alternating Current Stimulation (tACS) synchronised with sounds has been shown to modulate the perception of speech (1). However, the neural and perceptual mechanisms by which tACS changes speech perception remain undetermined. Degraded speech becomes more intelligible when signals are clearer, or when prior predictions are more accurate. These two factors have dissociable effects on neural activity (2,3). Experiments combining tACS with concurrent brain imaging and behavioural measures can determine which tACS-induced changes to brain activity directly cause changes in speech perception. By targeting more specific neural and perceptual processes we can help listeners that struggle to understand speech due to hearing or language impairments.

In the Digital Mental Health group, we are interested in how adolescent mental health outcomes are impacted by social media use in many different ways. There are therefore various potential avenues for a PhD in this space, and we love to hear your own original ideas. For inspiration, a non-exhaustive list of some example research areas can be found below (with further details on our group website https://www.orben.group/phd-study):

1) Development: How do developmental processes in adolescence, and potentially childhood, intersect with social media use and mental health? A PhD project could focus on how certain pubertal, cognitive or neural changes during this time impact how adolescents interact with and react to social media and other digital environments.

2) Cognition: What are the cognitive mechanisms linking social media use to mental health (either in clinical or community samples of adolescents)? A PhD project might, for example, take a particular element of social media use or design, match it with cognitive theories/paradigms and then use a range of methodologies to examine whether this change impacts mental health.

3) Computational Approaches: How can we apply analytical or computational methods from across the cognitive sciences to our research questions of interest? A PhD project might, for example, apply computational foraging models to social media behaviours, using data scraped from platforms or collected via data donation (see reference).

4) Assessing and Addressing Social Media Use in a Clinical Context: How does social media use a) predict adolescents receiving a mental health diagnosis, b) impact those living with a mental health diagnosis, or c) determine recovery? A PhD project could, for example, examine whether some concerns about social media use negatively impacting mental health are transdiagnostic (e.g., the ‘always on’ nature of social media) rather than disorder-specific. You could also study how social media use is assessed and examined in a clinical context, and how such processes could be improved and supported.

Working memory (WM) allows us to “actively” focus on a current thought while keeping a larger body of current background information “hidden”. Active WM has been studied extensively, and is often considered the core of WM. Despite its significance in computational models, the concept of background or “hidden” WM remains less understood. This is because hidden WM is marked by minimal neural activity making it difficult to see with neuroimaging. This project will explore the elusive nature of hidden WM by leveraging a cutting-edge approach that combines transcranial magnetic stimulation (TMS) with high-resolution functional magnetic resonance imaging (fMRI). By “pinging” hidden states—similar to how sonar pulses map the ocean floor—we aim to momentarily “reactivate” these hidden states. This project aims to (1) build a novel neural account for how the brain toggles between foreground and background information (2) advance precision neurotechnologies to offer unmatched resolutions for cognitive and clinical research.

This project explores cognitive control circuits by studying brain activity during video game play, a more natural and goal-directed behaviour than traditional lab tasks. Participants will engage in first-person/action video games while being scanned using advanced Human Connectome Project (HCP) imaging methods, allowing precise anatomical mapping. The research involves both exploratory analysis to generate new hypotheses and targeted testing of existing ones, offering hands-on experience in setting up the gaming system and integrating it with neuroimaging tools.

The “multiple-demand” (MD) circuit is a core brain circuit hypothesized to play a key role in cognitive control, yet our understanding of its functionality remains incomplete due to the limitations of non-invasive imaging in capturing its detailed anatomical connections. To address this challenge, this project aims to bridge advanced human brain imaging data with fine-grained anatomical data from non-human primates (NHPs). By working with existing invasive tract-tracing data from NHPs provided by collaborators, this research will utilize innovative cross-species tools to map and compare the MD circuit in both humans and NHPs, offering new insights into its detailed connections and functionality, ultimately bridging gaps between human and primate neuroscience.

While fMRI is invaluable in identifying the precise location of cognitive control circuits, its time resolution is limited to seconds while neural population activity operates at faster timescales. This project leverages rare data, provided by collaborators, in which intracranial electrodes are co-localized with patient-specific fMRI networks. Precision fMRI will allow delineating electrodes based on their circuit membership. ECoG will allow (1) tracking fast temporal exchanges between adjacent cognitive control circuits (2) use electrode stimulation to delineate the causal roles of distinct circuit components. The latter has immediate translational potential for the pre-surgical mapping of executive functions.

Yoga trains the brain to synchronise and build awareness of specific muscles for safe engagement, similar to rehabilitation processes where neural pathways are strengthened to restore function. It also serves as a valuable model for studying proprioception, motor skill acquisition, and expertise development. By integrating muscle recordings (surface electromyography (EMG) sensors) in realtime with biofeedback, our project aims to develop a technology to improve motor learning to engage specific muscle patterns, prevent injuries, and support muscle rehabilitation. We will use empirical testing to guide the parameters of biofeedback to achieve best motor training, by combining knowledge from cognitive and perceptual learning paradigms. We will iterate training with machine leaning to optimise learning between the technology and the human. This approach addresses key needs in musculoskeletal health while advancing our understanding of the neural mechanisms underlying motor control and expertise. Finally, partnering with yoga professionals, we will test the system’s efficacy in augmenting learning during yoga practice. Ultimately, this project aims to improve both physical and mental health by making yoga practice safer and more effective.