Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.

Training data of a machine learning application may be collected from geographically different places, but transferring distributed data to a central server for training can be legally or practically impossible. This has led to a fast-growing area in machine learning, i.e., federated learning. However, it was discovered that the federated learning in its original form, where a server iteratively aggregates local gradients received from the distributed users, may suffer from privacy leakage as an attacker can infer sensitive information from the local gradients. Since then, many improvements have been presented to mitigate this privacy leakage, but they often achieve better privacy with the compromise of other critical properties such as the accuracy and communication complexity. In this paper, we propose a novel solution that is based on a fully decentralised distributed learning. Our solution enables the optimisation over, and achieve a balance between, multiple properties, including privacy preservation, accuracy, communication complexity, efficiency, and tolerance to user failures. Our solution proceeds by first synthesising a communication topology between users according to the required properties and then applying a fully decentralised distributed learning where the server is not involved in the computation. In the decentralised learning, the aggregation of local gradients is reduced to a distributed consensus between users. Finally, the agreed value of the users is sent to the server after added a differential privacy noise. We are conducting experiments on both use cases in the US/UK privacy enhancing technologies challenge to validate our solution.

The ARISTOTELES project aimsto build a multinational harmonized data platform to develop and implement novel artificial intelligence (AI) approaches for management of complex diseases, where progression and manifestations of comorbidities are via multiple interacting pathways. We aim to apply our novel approach to a population of great need due to atrial fibrillation (AF), but our outputs can be extended to other complex diseases with multimorbidity. By integrating AIs into clinical practice, our platform will form a backbone for acceptable, responsible, and respectful uses of patient/participant data to develop and validate novel trustworthy AI tools for more personalized risk assessment and management. This represents a paradigm shift in AF treatment, moving from a focus on individual risk factors and selected outcomes (eg. stroke) to a holistic approach, underpinning timely diagnostic and therapeutic interventions to reduce disease progression, disability, hospitalizations and mortality, as well as improve patient adherence to lifestyle modifications, medications, and other treatment regimens. ARISTOTELES will be delivered through 8 inter-linking work-packages (WPs): WP1 is study management/coordination. WP2 provides the ethical/legal requirements for the development of a trustworthy AI. WP3 addresses stakeholder understanding of AI, needs assessment, and engagement in all the phases of the AI development. In WP4, granular data on genotype and phenotype characteristics, are harmonized from different datasets into a common platform. In WP5, AI algorithms/tools are developed and connected to an interactive output interface for patients and clinicians. In WP6, we test the AI tool developed in WP5 in a clinical trial simulation (in silico trial). In WP7 a multicenter randomized trial runs across multiple countries including both primary care and secondary care. WP7 and WP8 drive the clinical implementation and dissemination of results.

The ColdSpark project will validate a novel non-thermal plasma technology to produce hydrogen at an industrial scale from methane, with a process energy efficiency of 79%, achieving a conversion rate of 85% with zero CO2 emissions. This will be achieved by designing an industrial relevant reactor that leverages the best features of the non-thermal plasma technologies, gliding arc and corona discharge, to ensure high efficiency and scalability. The innovation addresses for the first time the critical step of matching the reactor with a pulsed power supply. It enables a perfect fine-tuning of the cracking process parameters, to find the right electron density and energy distribution in the plasma reactor, to maximise energy efficiency. The up- and downstream gas management will be optimised to further contribute to the system’s compatibility to existing infrastructure. The project will develop and test a novel plasma reactor at lab scale and validate it in conjunction with the power supply at large-scale, pursuing the industry’s most power efficient generation of hydrogen alongside high-value carbon. The technology will assess its application for both, natural gas and biomethane producers. A low energy cost (< 15 kWh/kg H2 produced) without the need for catalysts and water, makes the proposed solution the most cost-competitive, environment friendly, and less complex to implement. The reactor design and modularity bring lower CAPEX and OPEX and make it easily scalable and flexible. The project gathers the expertise of a mix of academic, research, and industrial partners from five countries, which bring both outstanding research and topic competence, as well as knowledge and access to the solution for end-user industries.

According to the WHO Osteoarthritis (OA) is one major course of years lived with disability in the elderly and is considered a high-burden disease, which makes it a research priority in Europe. There is no cure for OA and SoA treatments need to be reconsidered. Current pharmacological interventions consist of analgesic, and anti-inflammatory drugs as well as intraarticular steroids and hyaluronic acid (IA-HA) with moderate efficacy and associated long-term side effects. New medications are thus needed both to alleviate pain and slow down disease progression. Taking advantage of the explosion of RNA technologies in the last years, SINPAIN aims to develop a pipeline of siRNA-based therapy built on the combination of current technologies (dynamic IA-HA and nanocarriers) that will be designed step-by-step in order to reach successful management of inflammation and innervation therapy for the treatment of early (grade 0-1) and later stages (grade 3-4) of knee osteoarthritis (OA). To do so, a nanoformulation composed of functional IA-HA that can be loaded with vectors, for the delivery of siRNA targeting IL1ß and NGF, and nanocarriers will be developed. In parallel, a large effort will focus on understanding the pathological mechanisms of OA. To validate efficacy in relevant potency assays, 3D coculture models will be developed with human cells and tested in unique bioreactors mimicking joint environments and biomechanics. With the identified cell targets, IA-HA will be modified with an immunomodulator peptide which will activate the adaptive immune response, responsible for OA regeneration. The 4 pipeline products of SINPAIN will be validated in vivo in a relevant OA model with SoA techniques that will demonstrate the reduced inflammation and pain, as well as the cartilage regeneration for the last product. Taking advantage of all the data obtained during the project, a decision-making tool based on machine learning will be validated to offer patients a personalized therapy.