The increasingly higher medical care costs have reduced the access to a substantial percentage of the population, and this could come with dramatic consequences for the entire humanity. In this context, it is clear the need for precision and predictive approaches are needed to both reduce the cost per capita and design personalised treatments. As a composite ensemble of useful IoT devices, an AI-based patient digital twin represents a sort of avatar of the patient, gathering data through a composite ensemble of useful IoT (Internet of Things) devices. By combining an AI-based patient digital twin with game-theoretic modelling approach in multiscale environments and deep learning methodologies, this PhD project aims to explore the evolutionary dynamics of personalised intervention strategies and design cost-effective incentive mechanisms and personalised decision support systems. It will allow reducing the healthcare costs and increase the explainability in the deep learning analysis of patient data.
Background
Nowadays, one of the main challenges in medicine is represented by the medical care costs and most of humanity cannot afford even basic medical care due to the growing costs that have been reducing the access to a substantial percentage of the population. To reduce healthcare costs, we need a dramatic paradigm shift in healthcare passing from a reactive to a proactive or predictive approach in medicine. The idea is therefore to leverage the emerging concepts of precision and predictive medicine, exploiting the technology-driven digital transformation of the health service and the multiscale integration of large amount of data at both individual and population levels. Since these data will be taken at different scales, from genomic data to clinical data at systemic levels, designing AI-based, machine learning and, more specifically, deep learning methods, can allow us to achieve this target of integrating multiscale data. In terms of data, one of the main issues is given by the reliability of data related to the patients, indeed reliable systems accurately forecasting physiological conditions of patients is one of the primary targets of precision medicine. The underlying complexity of human physiology, where some systems are independent from others, and the need for integrating and analysing different types of data makes this research field extremely challenging. Considering multi-omics data, this complexity is even more marked if we consider the interplay between communities (i.e., microbial communities) at a meso-scale level. Furthermore, the complexity is not only at a micro-scale level, in terms of multi-omics data, within the human body, but also at a meso-scale level and macro-scale levels, between communities of agents represented by different data gathered from multiple digital twins.
Therefore, to disentangle this complexity, preliminary work has showed the importance of a multiscale analysis and modelling approach that embodies both macro-scale, meso-scale, and micro-scale aspects (in system biology, multiscale means considering from DNA to RNA to protein, from metabolites to cells to tissues, from tissues to organisms and even to interacting populations, and microbial communities). To this aim, a meaningful approach is the integration of reliable multiscale data and models able to combine these data by using deep learning methodologies and build personalised decision support systems and incentive mechanisms able to make the entire healthcare system cost-effective and more explainable.
Methodology
This research will be built upon methodologies and outcomes from the proposed direct of research’s past and ongoing research works and collaborations within the SCEDT, and with other Universities, such as University of Cambridge, where we are working on “Managing post covid hospital waiting times using digital twin and game theory”, with Professor Pietro Liò, Full Professor at the department of Computer Science and Technology of the University of Cambridge and a member of the Artificial Intelligence group and the Cambridge Centre for AI in Medicine. The following are some key publications in the areas of game theory, complex networks and network science: PloS one, 10(10), e0140646; PLoS computational biology, 15(1), e1006714; IEEE Transactions on Cognitive Communications and Networking, doi: 10.1109/TCCN.2020.3027697; Elsevier, Pervasive and
Mobile Computing, https://doi.org/10.1016/j.pmcj.2020.101230; Scientific reports, 6, 37105; deep learning; Spasov, S. et al., 2021; and explainability
Main aim
To design and explore the evolutionary dynamics of personalized intervention strategies and cost-effective incentive mechanism, by combining an AI-based patient digital twin with Game Theory in multiscale environments and Deep Learning methods to increase the explainability in the deep learning analysis of patient data. The final target is to address the abovementioned challenges and assumptions, and therefore generate new insights in the field of personalized and precision medicine. To achieve this goal, we shall combine research on incentives and multiplex network modelling with the data integration from AI-based patient digital twin, deriving dynamical approaches to analyze the evolutionary dynamics of behaviours of these agents as digital twins in a complex multiscale environment. This modelling approach will enable us to systematically explore how the self-awareness of the patients and different types of multiscale incentives might influence evolutionary behaviors and improve healthcare decisions.
Research strength and contributions to the school strategy
This project will contribute to the school growing international reputation in behavioural modelling research, incentives design, and AI-related research, enabling these important research areas to become a Teesside and SCEDT research identity, both at a national and international level. Medical care growing costs and the reduced access to healthcare of a substantial percentage of the population represents a big concern. This PhD project will bring fundamental contributions that help us better understand how to reduce healthcare costs, and build precision and predictive medicine approaches. These will allow us not only to design cost-effective incentive mechanisms and personalised decision support systems, but also to increase the explainability in the deep learning analysis of patient data. Given the research community’s extensive engagement with healthcare costs, AI-based tools and methodologies, and their emphasis in the UK National AI Strategy, this project is timely, topical and of potentially great practical importance.