Unified probabilistic modelling of adaptive spatial-temporal structures in the human brain

Learning from experience and adapting our behaviour to new situations is a fundamental skill for our everyday interactions. But what are the brain plasticity mechanisms that mediate an individual's ability to make progress during training on complex tasks? What is it that differentiates `good' from `poor' learners in their ability to adapt? Recent advances in functional brain imaging technology provide us with the unique opportunity to study how the human brain changes with learning. However, the existing methods focus predominantly on modelling brain activity data within a single session rather than across training sessions. As such, these methods are not capable of capturing larger scale dependencies emerging in brain activity as training progresses. We will develop a novel methodology that allows holistic unified modelling of a series of brain imaging data measured during the course of learning. Using this methodology we will study brain changes that result from extensive training on complex visual tasks. Our work will offer scientists and practitioners advanced tools for using brain activity measurements to understand the brain learning mechanisms and how they improve our ability to make complex decisions. The proposed methodology may have predictive power for making inferences about 'prototypical' learning patterns that can be used to predict adaptive behaviour in individuals with different learning strategies and design training schemes tailored to the individuals' abilities and needs. Hence our findings have potential implications for the design of dedicated training programmes that take into account an individual's learning capacity. Such programmes may have applications in education or intervention and rehabilitation in normal and pathological development and ageing (e.g. stroke, neurodegenerative disorders).

Conventional methods for the analysis of functional MRI data have been instrumental in revealing the human brain architecture, allowing us to identify cortical areas involved in processing sensory information, as well as circuits engaged in higher cognitive functions. However, characterising the link between activity in these networks and adaptive behaviour remains an open challenge in cognitive neuroscience. In particular, understanding how the brain circuits and functions change with experience is fundamental for unravelling the plasticity of the human brain. Yet, we are still lacking principled methods for modelling and analysing in a unified framework the dependencies of brain activity across stages of training (within and across multiple sessions) that mediate our ability to make progress in learning complex tasks. We propose to develop a novel methodology based on generative probabilistic modelling for holistic analysis of complex brain imaging data measured during multiple sessions and across stages of training. The model is formed by a set of coupled hidden process models, one for each session. To stabilise the parameter fitting and enhance interpretability of the model, the hidden process models will be constrained by 1) probabilistic pooling of voxels into spatially contiguous groups of common parametrised hemodynamic response; 2) reducing the uncertainty in the onset time of hidden processes by concurrent fMRI and EEG measurements. We will test and validate our methodology in a large set of controlled experiments using synthetic data, as well as real data measured in participants trained to perform complex visual decision tasks. The proposed work will provide advanced methodological tools for studying the human brain plasticity mechanisms that mediate our ability to train and improve in complex tasks and novel theoretical insights into the functional plasticity signatures of learning in the individual human brain.

Most of the current functional brain imaging research is concerned with understanding neural mechanisms involved in cognitive processes as revealed by single session scans. However, investigating neural mechanisms in the human brain that mediate our ability to train and improve in complex tasks requires a new set of tools capable of capturing dependencies in the brain imaging data across multiple sessions characterising different stages of learning. The proposed project will deliver a novel methodology for a unified systematic processing and analysis of such complex data. The work has a strong potential for building and enhancing interdisciplinary, high-end research in cognitive neuroscience in the UK. This basic research on learning and brain plasticity has the potential to improve the quality of life through applications in public health, education and policy. Who will benefit from this research? The main beneficiaries outside the academic community will be the general population engaged in everyday activities entailing training in complex tasks as well as individuals with learning difficulties of different kinds, professionals from health and social services working with them and policy makers. Further, we will exploit opportunities for applied research and knowledge transfer: companies developing advanced tools for brain imaging research, designing educational software or brain training games. How will they benefit from this research? Our findings will advance our understanding of the link between activity in cortical areas involved in the processing of sensory information, as well as circuits engaged in higher cognitive functions (attention, memory, decision making) and adaptive behaviour. Delineating brain learning mechanisms is critical for the design of learning programmes for education or intervention that are tailored, accessible and beneficial to people of all ages and conditions. What will be done to ensure that they have the opportunity to benefit from this research? We will disseminate our work to the public, third and private sector through: 1. project-specific website and links to collaborators' sites 2. keeping participants in our research informed about findings and applications 3. the University and BBSRC Press Office (e.g. National and local media, Science Festival). 4. activities for the public organised by the University and the Alumni Office: presentations at Open Days, lay-reports for magazines. 5. Workshops through the Bridging the Gap programme engaging scientists, practitioners, participants in our studies, and the general public.