AI-driven brain modelling for personalised cognitive enhancement

Lead Research Organisation: University of Bristol
Department Name: Computer Science

Abstract

Cognitive decline is a core feature of dementia and mild cognitive impairment (MCI) which combined affect over 1.3 million people costing the UK over £35 billion. There are cognitive-enhancing drugs licensed to alleviate cognitive decline in dementia and MCI. However, they provide only a modest gain to around half the people who take them, and we do not understand how to make these drugs more efficient.
Artificial intelligence (AI) neural networks have shown remarkable ability to stratify patients and predict susceptibility to brain disorders, such as dementia. However, the solutions of these algorithms are often hard to interpret and inform new research and treatments. To address this, we have recently started mapping AI neural networks onto biological neural networks. In contrast to standard approaches, our AI-driven brain modelling approach enables us to map precise neuronal mechanisms onto cognitive processes in health and disease. Our models predict that specific neuromodulators enable the development of memories in biological neural networks with heightened resilience to perturbations. In this project we will use our AI-driven modelling to study how neuromodulator malfunction exacerbates the decline in cognitive function that results from neuronal death. Consequently, our model will generate specific cognitive training treatments that are optimal for reducing cognitive decline during dementia and MCI. Next, we will simulate commonly used cognitive-enhancing drugs to study why current drugs only offer modest improvements and how to best combine them with cognitive training. To perform this work, we will bring together a team of leaders in computational, fundamental, and clinical neuroscience with AI experts.
Overall, this project will develop the first AI-driven brain modelling tool capable of facilitating cognitive enhancement research. This line of AI-informed research will lead to personalised treatments that will improve quality of life in dementia and MCI.

Technical Summary

Drugs that target the cholinergic system, such as donepezil and galantamine, are commonly used to alleviate cognitive decline in dementia and mild cognitive impairment (MCI). However, these drugs only provide a modest benefit in a subset of patients. Artificial intelligent algorithms are showing high promise in predicting and classifying brain disorders, including dementia. However, they are typically used as black box approaches which fail to provide insights on how biological mechanisms should be rescued to improve clinical outcomes.
Recently we have developed AI-driven brain models that offer great promise at bridging the clinical and fundamental neuroscience gap. In contrast to standard approaches, our AI-driven brain modelling approach enables us to map precise neuronal mechanisms onto cognitive processes in health and disease. Our models predict that Cholinergic neuromodulators enable the development of neuronal task-specific representations in biological neural networks with improved resilience to perturbations, such as neuronal death.
In this project we will study how the cognitive robustness to neuronal death predicted by our model depends on cognitive training conditions. This will enable us to identify the optimal training regimes for improved cognitive robustness. Next, we will build on these findings together with a model of cholinergic drugs (donepezil and galantamine) to propose how to best combine cognitive training with drug administration, and why under some conditions cholinergic drugs might not be efficient. These results will guide the development of personalised treatments for cognitive enhancement through a combination of existing cholinergic drugs with cognitive training. Overall, this project will showcase how AI-driven brain modelling closely informed by decades of experimental research can inform clinical neuroscience.

Publications

10 25 50