Virtually all of the Cam-CAN analyses, both neuroimaging and behavioural, are ultimately aimed at integration across domains and modalities to fully understand the ageing process. See pages on functional analyses for further discussion of cross-modal analyses using fMRI and MEG
A number of tools and methods are currently under development for use in integrating across modalities in the Cam-CAN project:
1. Structural Equation Modelling (SEM) toolbox
Structural Equation Modelling (SEM) is a methodological framework to test hypotheses about how a set of variables relate to each other. In cognitive neuroscience, it can be used to test hypotheses about brain-behaviour relationships. Since existing SEM packages are not available within a MATLAB environment, we developed a MATLAB-based SEM toolbox. This will make the analysis seamless, since most other analyses for the project will also be done in MATLAB.
The toolbox allows for model specification, model estimation and model validation. It can be run through an easy-to-use GUI or in MATLAB batch-mode. The figure on the right shows the GUI:
A useful feature of the SEM methodology is that it naturally allows to test hypotheses about how variables of different types and from different imaging modalities relate to each other. Further work will include fitting models to data from different cognitive tests. The range of neuro-imaging variables used will also be expanded to include novel structural variables (eg. Mean Diffusivity, Fractional Anisotropy) as well as functional variables like BOLD activity, ERF amplitude, etc.
2. Mediation models
A method closely related to SEM is that of mediation and moderation models. Age-related declines in many cognitive functions are well established; for example older adults demonstrate declines in episodic memory or language production. These declines may be mediated by other age-related changes, such as general-purpose cognitive processes or structural changes in the brain (e.g. decrease in white matter (WM) integrity connecting brain regions). Current analyses are examining the ways in which age-related cognitive decline is mediated by general cognitive measures, by neural changes, and by cardiovascular fitness (CVF).
For example, we are examining whether WM integrity and CVF could be influencing cognitive decline independently (see figure below, panel A). In order to address this question, we used mediation analysis, which indicated that age-related declines in fluid intelligence (see figure below, panel C) are mediated by white matter quality in a number of prefrontal cortical (PFC) regions. We first performed a whole-brain voxel wise analysis of diffusion tensor imaging (DTI) data to identify regions of cortical WM that mediate age-related declines in fluid intelligence (see figure below, panel B). We then applied these values in our conceptual mediation model, along with measures of CVF (systolic blood pressure). This model (see figure below, panel D) demonstrated that WM had a greater mediating influence than CVF, and furthermore that the model overall attenuated the relationship with cognitive decline predicted by age alone. The aim of this approach is to demonstrate the relative contributions of white matter health and cardiovascular fitness in predicting individual differences in age-related cognitive decline.
3.Movie data: integrating across multiple cognitive domains
Cam-CAN participants who have MR scans also watch an excerpt of a movie to provide information on neural activity in a more naturalistic setting. A good director makes movies that are highly engaging, and as a result, activate a broad range of brain regions. The pictures and sound stimulate the auditory and visual system; watching people move or talk activates the motor system; thinking about the intricacies of the plot activates regions in the frontal and parietal lobe; and so on. This broad activation provides us with a way to assess many brain regions within one short scan. We can then look at what is common about the activation patterns across people; and what changes through the lifespan. We can examine what aspects of the brain’s response predict performance on another task (e.g., a test of memory).Thus, a good Director also produces patterns of brain activity that to the scientist are compelling viewing.