Principal Investigator: Dr Parashkev Nachev
Department: Institute of Neurology
Institution: University College London (UCL)Tags: 16273, brain-body interactions, lifestyle, Machine Learning
1a: Lifestyle is thought to contribute to most major illnesses. So complex and diverse are our lives, however, that the important factors are difficult to determine without totalitarian levels of supervision. Since the behavioural traits that strongly influence our lifestyles are mirrored in the individual architecture of the brain, we propose instead to link lifestyle-relevant features of the brain to physiological markers predictive of disease, assisted by “artificial intelligence”-based modelling techniques capable of dealing with such complex problems.
1b: A successfully completed project would isolate brain “signature patterns” of behaviours predictive of bodily markers associated with the major illnesses ischaemic heart disease, stroke, diabetes, COPD, Parkinson’s disease, and mood disorders. Such signatures could help, a) to understand the biological basis of such behaviours, b) to predict behaviour before it has had physical consequences so as to focus disease prevention, and c) identify neural targets for potential pharmacological or neurophysiological modulation in the future. The research therefore falls within UKBiobank’s stated purpose to the extent of assisting the prevention, diagnosis, and potentially treatment of the major illnesses cited.
1c: The research has three components: a) analysis of magnetic resonance brain imaging together with behavioural/cognitive markers so as to isolate neural signature patterns predictive of characteristic behavioural traits. b) analogous analysis of brain imaging together with bodily markers associated with the illnesses of interest, both directly measured and derived from body imaging. c) statistical quantification of the predictive power of models incorporating joint information from the brain-behaviour and brain-body signature patterns. The final output will be signature brain maps for bodily markers of interest, quantified by sensitivity and specificity measures based on the whole dataset.
1d: We are seeking to identify complex biological patterns that are individually predictive with sufficient precision to be of potential clinical use. This necessitates not resolving between crude groups of people but between many clusters of people sharing arrays of many features. These clusters will be too diverse for the power to do this to be easily derived from any sample. The most appropriate approach therefore is to analyse the full cohort for which brain imaging is available.