Potential PhD/MPhil topics

Our research generally concerns the cognitive neuroscience of ageing and dementia. This involves relating various neuroimaging measures of the brain to various latent cognitive abilities, as well as lifestyles, to understand how ageing and neurodegenerative diseases like Alzheimer's impair cognition through affecting the brain, and how to mitigate these effects. We tend to use large open-access datasets like the Cambridge Centre for Ageing and Neuroscience (Cam-CAN; see recent review), BioFIND, NTAD/SHINE, UK BioBank, cohorts in the LifeBrain consortium and those on the DPUK portal, combined with advanced statistical techniques.

To hear an overview about our work on healthy ageing using the CamCAN cohort (Jan 2021), you could watch this video.

Potential MPhil Projects

Note these projects have data collected already from large cohorts, so the projects focus on data analysis and can begin immediately. Some PhD students may also be willing to assist with project supervision.

1. Disentangling Episodic Memory from General Cognition: a Longitudinal Modelling Framework

Is memory its own thing, or just a part of general intelligence (g)? Our previous cross-sectional work in Cam-CAN showed memory carries unique age-related variance beyond fluid intelligence (g), supporting the idea that decline in memory is a distinct process from decline in intelligence (https://www.nature.com/articles/srep32527). A much stronger test, however, is longitudinal (see https://doi.org/10.1037/a0024503 on the importance of longitudinal data): do memory and other cognitive abilities follow different trajectories over time as people age? Using large-scale longitudinal ageing datasets (e.g. Cam-CAN, OASIS-3, HABS-HD, DLBS), one can identify the factor structure of neuropsychological batteries that are routinely administered in clinics, and test if these factors can be reliably tracked over time. To do so, we will use advanced statistical models, such as longitudinal extension of exploratory structural equation modelling (ESEM), and other techniques (e.g. latent change score modelling, LCSM) (see https://github.com/RikHenson/AgeBrainCognition or https://stats.oarc.ucla.edu/r/seminars/rsem/ for a brief introduction on SEM). This project requires mainly statistical competence, but can dramatically expand knowledge of field.

2. How does memory decline relate to hippocampal structure/function and vascular health?

Using the CamCAN longitudinal structural MRI, functional MRI and story memory data in N~130 people (and possibly other open access cohorts we have downloaded), we can 1) try to replicate relationship between hippocampal volume change and memory change, possibly only in APOE e4-carriers (https://doi.org/10.1002/dad2.12110) and 2) test whether resting-state fluctuation amplitudes (RSFA) from fMRI capture any more change-change variance, with the hypothesis that hippocampus is particularly sensitive to age-related declines in vascular supply (which could be further tested by moderations by pulse pressure, https://doi.org/10.1161/HYPERTENSIONAHA.124.24543). Since tabular values already extracted, this project requires mainly statistical competence, but can dramatically expand knowledge of field.

3. Can we estimate general “excitability” of the brain using fMRI and MEG?

Simple models relate the Excitatory:Inhibitory (E:I) ratio to the “1/f” aperiodic slope of M/EEG resting-state activity (eg, https://doi.org/10.1002/dad2.12477) and to the covariance of fMRI timseries (eg, https://www.science.org/doi/10.1126/sciadv.adr8164). In the Cam-CAN dataset, we have both MEG and fMRI on ~700 people at baseline (plus ~130 of these 12 years later) so can compare these estimates, and explore as a function of age. This project requires programming ability (to estimate E:I values) and basic familiarity with MEG and fMRI, but can dramatically expand knowledge of both techniques.

4. How do age-related changes in physical activity affect cognitive and brain changes?

Using the CamCAN longitudinal data in N~130 people over ~12 years, we can test change-change relationships between these variables, particularly measures of white-matter (see https://doi.org/10.1093/gerona/gly220 for preliminary, cross-sectional work). This introduction to statistical modelling of ageing effects may help: https://github.com/RikHenson/AgeBrainCognition. However, it is likely that larger/replication samples will be needed, which will involve identifying other open-access longitudinal cohorts with relevant variables. This project requires mainly statistical competence, but can dramatically expand knowledge of field. A sub-project could use a small validation dataset with 4 T1 scans per person to estimate reliability of hippocampal volume, e.g., from SamSEG versus FreeSurfer.

5. Is resting-state MEG useful to detect early dementia?

We have collected some resting-state MEG datasets (BioFIND, NTAD/SHINE) from which various features can be extracted for machine-learning techniques to classify MCI vs Controls, or MCI-to-AD conversion (eg, https://doi.org/10.1016/j.neuroimage.2022.119054). For example, the 1/f aperiodic slope of MEG activity has been related to Excitatory:Inhibitory ratio (E:I) in the brain (see Other Project), and Amyloid/Tau build-up seem to have different effects on this E:I ratio (https://doi.org/10.1002/dad2.12477). Using BioFIND and NTAD/SHINE datasets, we can test this hypothesis by comparing MCI convertors/non-convertors/AD biomarker-confirmed patients with matched controls, and/or as a function of disease progression. This project requires programming ability (to extract MEG features) and basic familiarity with MEG or EEG, but can dramatically expand knowledge of MEG and dementia. This talk may help: https://youtu.be/pKieBZhBKg0

6. Is an implicit repetition MEG paradigm useful to detect early dementia?

We have developed an MEG task using incidental repetition of visual scenes to index the brain's automatic ability to "remember" stimuli, independent of strategy, motivation, etc, which we expect to be sensitive to early neurodegeneration, eg in Alzheimer's Disease (AD). We believe this effect depends on medial temporal lobe structures that are often affected earlier in AD, even if "read-out" of the repetition with MEG arises from connected cortical regions. We have collected data from a very large number of participants across several cohorts (healthy older people in "Cam-CAN", early MCI in "Deep & Frequent Phenotyping" and confirmed early AD in "NTAD/SHINE"; see https://doi.org/10.1136/bmjopen-2021-055135). Initial analyses revealed only a weak, late repetition-related ERF, but more sophisticated analyses (eg multivariate decoding of time-frequency data across sensors/sources) may reveal other repetition-related effects that could be used to detect and track neurodegeneration. This project requires programming ability (to extract MEG features) and basic familiarity with MEG or EEG, but can dramatically expand knowledge of MEG and dementia.

7. Task-general functional compensation in ageing

We have explored the possibility that older people can demonstrate increased functional activation (or connectivity) during performance of cognitive tasks in order to compensate for age-related structural loss. Until now, we have failed to find convincing evidence that the age-related hyper-activation in frontal (https://doi.org/DOI:10.1523/JNEUROSCI.1701-17.2018) or lateral (https://doi.org/10.1523/JNEUROSCI.1111-21.2021) cortices is compensatory (rather than just inefficiency). Only in a problem-solving matrix task have we found activation that meets criteria for being compensatory (https://doi.org/10.7554/eLife.93327.1), though this was in a posterior visual region, and may be specific to this particular task. We would now like to examine if there is a task-general pattern of activation/connectivity that is compensatory, using a range of fMRI data in Cam-CAN and possibly in a cohort called RANN in collaboration with US colleagues.

8. Affective representations during the intentional regulation of emotional responses

Previous research has demonstrated that activity patterns in the orbitofrontal cortex (OFC) reflect the subjective experience of valence (https://www.nature.com/articles/nn.3749?). Simultaneously, the OFC interacts with the medial prefrontal cortex and the amygdala which form a central network responsible for the regulation of emotions (https://www.nature.com/articles/s41380-022-01883-2). Does the OFC track or shape emotional experience to a similar extent during natural affective experience and intentional emotion regulation? This project will use the CamCAN emotion reactivity and regulation (ERRT) fMRI task data to investigate whether subjective ratings of stimulus valence during short movie scenes are reflected in OFC and limbic BOLD patterns, depending on the explicit instruction to regulate the emotional response. Insights based on this project may advance our mechanistic understanding of the emotion regulation network and thus potentially have important implications for the treatment of affective disorders.

Potential PhD Projects

Some brief suggestions are below, but note that the list below is not exhaustive; i.e, if you have a topic in mind that overlaps with our general interests, please do email to discuss! Note suggestions are deliberately brief since part of our application process is to establish potential to develop independent ideas.

1. Investigating the neural correlates of "cognitive reserve"

How can some people maintain cognitive performance into late-life despite clinical MRI evidence of typical age-related brain atrophy (e.g., Henson, 2026), such as functional segregation of brain networks (Raykov et al., 2023)

2. Determining the lifestyle contributions to "cognitive reserve"

Some examples include effects of mid-life activity (Chan et al., 2018), as covered in this BNA Festive talk.

3. Investigating functional compensation while older adults perform cognitive tasks

We currently believe evidence is currently limited, at least for memory (Morcom et al., 2018) and motor control (Knights et al., 2021), but possibly occurs in problem-solving tasks (Knights et al., 2024), though this may reflect task-specific strategies, so common mechanisms need to be explored across multiple tasks.

4. Importance of vascular health for late-life cognition

This includes effects of variable like pulse pressure on white-matter, such as in this paper (King et al., 2023) and as summarised in second half of this talk.

5. Behavioural, fMRI and/or M/EEG studies on the effect of healthy ageing on memory

Continuing work such as (Henson et al., 2016) or (Gellersen et al., 2022).

6. Investigation of MEG for early detection of dementia

...particularly using machine learning classification methods, as in (Vaghari et al, 2022), summarised in this talk.

7. Investigation of the "cognitively frail"

These are people who score below cut-offs on conventional tests for dementia (such as MMSE/ACER) yet have not sought clinical advice. We have only reported preliminary MEG data in the Cam-CAN frail cohort (Kocagoncu et al., 2022), but have a lot more data to examine.

For other possible projects using CamCAN data, see https://www.youtube.com/watch?v=8f5B9tJe9BA.