Selected Projects
Brief summary of selected projects
I this project I tested how brain networks re-organise across wide age span. I computed functional connectivity from 650 people aged 18-89 from the Cam-CAN cohort and examined whether older adults exhibit less modular brain networks. Specifically, I computed how brain regions from the same network (e.g., default mode network, attentional network) are connected among each other compared to regions from other networks. This within vs between network connectivity can be summarised by a measure of functional system segregation (SyS), and is though to represent how modular a brain network is. A modular functional network architecture may be important for optimal brain function and metabolic efficiency. Across 3 task states and different brain parcellations, I showed that SyS declined in older adults and predicted fluid intelligence, after accounting for age effects. Interestingly, SyS did not seem to mediate the positive effects midlife activities have on late-life cognition.
Apolipoprotein E ε4 is a major genetic risk factor for Alzheimer’s disease, and some apolipoprotein E ε4 carriers show Alzheimer’s disease–related neuropathology many years before cognitive changes are apparent. Therefore, studying healthy apolipoprotein E genotyped individuals offers an opportunity to investigate the earliest changes in brain measures that may signal the presence of disease-related processes. Here, we use a naturalistic activity (movie watching), and a marker of episodic memory encoding (transient changes in functional magnetic resonance imaging activity and functional connectivity around so-called ‘event boundaries’), to investigate potential phenotype differences associated with the apolipoprotein E ε4 genotype in a large sample of healthy adults. Using Bayes factor analyses, we found strong evidence against existence of differences associated with apolipoprotein E allelic status. Similarly, we did not find apolipoprotein E-associated differences when we ran exploratory analyses examining: functional system segregation across the whole brain, and connectivity within the default mode network. We conclude that apolipoprotein E genotype has little or no effect on how ongoing experiences are processed in healthy adults. The mild phenotype differences observed in some studies may reflect early effects of Alzheimer’s disease–related pathology in apolipoprotein E ε4 carriers.
In this collaborative project, I used 11 different white matter measures computed from diffusion weighted and non-diffusion weighted imaging available the Cam-CAN cohort. I performed factor analysis on these different measures of white matter integrity and found that 4 latent factors were enough to explain ~90% of the variance in all 11 WM measures. I examined how these latent white matter integrity measures related to cardiovascular health and cognition factors. I found that pulse pressure was associated with the latent white matter factor strongly loading on free-water content and white matter lesions. This is inline with current views that cardiovascular health is associated with health of white matter tracts. Interestingly, I also found that distinct latent white matter factors predicted distinct aspects of cognition such as fluid intelligence and processing speed, further suggesting that using latent factors rather than individual measures may not only provide statistical benefits to analysing brain-cognition relationships, but also increase biological interpretability of results.
In this project I tested how participants’ knowledge that events have beginnings, middle and ends affects their memory later on. Specifically I showed participants clips interrupted at different points in time. Some clips were interrupted just before an action was seen as completed (incomplete - blue); The complete clips (orange) showed the completed action; In the updated condition (purple) participants saw the completed action followed by a novel unrelated scene. We predicted that participants would make false memory errors about the incomplete clips because their memory would be biased by our prior experiences with complete events that have beginning, middle and end. Furthermore we expected that participants would need to build a detailed enough event model initially to make these false memory errors. Across five experiments we found that participants are more likely to make false memories for the ending of incomplete videos.
In this project, I was interested to understand what information do we actually represent when we remember complex events. Imagine you are trying to remember a meeting you had with Ben and Monika, last Tuesday. You often have meetings with Ben, but this was your only meeting that involved Monika. While trying to remember this meeting would you represent all elements, such as the conversation, the location, Monika, and Ben equally or would you weight some features for the situation more to help you differentiate this memory from other similar memories (e.g., other meetings last week). To address this question we asked participants to learn events involving people that were either unique or shared across events (e.g., Monika was present in one situation, but Ben is present in our memory of multiple situations). Participants then remembered the events within a fMRI scanner. We used machine learning to examine whether when people were remembering the particular situations we could decode information about the people involved in the memory. We found evidence that participants mainly reinstated features that are shared across events rather than showing evidence for holistic retrieval.
It is still unclear how we utilise our prior (schema) knowledge to learn novel information and which brain regions support this process. In this project, I trained participants with different TV shows to teach them a person schema. I then scanned participants while they watched and later remembered video clips taken from the trained and untrained show. I used pattern analysis to test whether participants showed similar patterns of brain activity in medial prefrontal cortex (MPFC) and posterior medial cortex (PMC) while watching the clips from the trained show, as suggested by theories of schema processing. I found that shared patterns of activity among clips of the trained show that predicted better memory for those clips. This was evidence that MPFC represents schema knowledge while encoding novel clips congruent with prior knowledge.
