Current hearing aids suffer from two major limitations: 1) hearing aid audio processing strategies are inflexible and do not adapt sufficiently to the listening environment, 2) hearing tests and hearing aid fitting procedures do not allow reliable diagnosis of the underlying nature of the hearing loss and frequently lead to poor fitting of devices. This research programme will use new machine learning methods to revolutionise both of these aspects of hearing aid technology, leading to intelligent hearing devices and testing procedures which actively learn about a patient's hearing loss enabling more personalised fitting. Intelligent audio processing The optimal audio processing strategy for a hearing aid depends on the acoustic environment. A conversation held in a quiet office, for example, should be processed in a different way from one held in a busy reverberant restaurant. Current high-end hearing aids do switch between a small number of different processing strategies based upon a simple acoustic environment classification system that monitors simple aspects of the incoming audio. However, the classification accuracy is limited, which is one of the reasons why hearing devices perform very poorly in noisy multi-source environments. Future intelligent devices should be able to recognise a far larger and more diverse set of audio environments, possibly using wireless communication with a smart phone. Moreover, the hearing aid should use this information to inform the way the sound is processed in the hearing aid. The purpose of the first arm of the project is to develop algorithms that will facilitate the development of such devices. One of the focuses will be on a class of sounds called audio textures, which are richly structured, but temporally homogeneous signals. Examples include: diners babbling at a restaurant; a train rattling along a track; wind howling through the trees; water running from a tap. Audio textures are often indicative of the environment and they therefore carry valuable information about the scene that could be harnessed by a hearing aid. Moreover, textures often corrupt target signals and their suppression can help the hearing impaired. We will develop efficient texture recognition systems that can identify the noises present in an environment. Then we will design and test bespoke real-time noise reduction strategies that utilise information about the audio textures present in the environment. Intelligent hearing devices Sensorineural hearing loss can be associated with many underlying causes. Within the cochlea there may be dysfunction of the inner hair cells (IHCs) or outer hair cells (OHCs), metabolic disturbance, and structural abnormalities. Ideally, audiologists should fit a patient's hearing aid based on detailed knowledge of the underlying cause of the hearing loss, since this determines the optimal device settings or whether to proceed with the intervention at. Unfortunately, the hearing test employed in current fitting procedures, called the audiogram, is not able to reliably distinguish between many different forms of hearing loss. More sophisticated hearing tests are needed, but it has proven hard to design them. In the second arm of the project we propose a different approach that refines a model of the patient's hearing loss after each stage of the test and uses this to automatically design and select stimuli for the next stage that are particularly informative. These tests will be be fast, accurate and capable of determining the form of the patient's specific underlying dysfunction. The model of a patient's hearing loss will then be used to setup hearing devices in an optimal way, using a mixture of computer simulation and listening test.

This project brings together unique expertise in Computational and Experimental Fluid Dynamics, Model Reduction and Artificial Intelligence, to identify solutions for the management of people and spaces in the current pandemic and post lockdown. A new interactive tool is proposed that evaluates the risk of infection in the indoor environment from droplets and aerosols generated when breathing, talking, coughing and sneezing. This capability will become more critical as winter approaches and building ventilation will need to be limited for comfort considerations. The fluid dynamic behaviour of droplets and aerosols, the effect of using face masks as well as other parameters such as room volume, ventilation and number of occupants are considered. A datahub capable of storing, curating and managing heterogeneous data from sources internal and external to the project will be created. A synergetic experimental and numerical approach will be undertaken. These will complement the existing literature and data from other EPSRC-funded projects providing suitable datasets with adequate resolution in time and space for all the relevant features. To support experiments and numerical simulations, reduced order models capable of interpolating and extrapolating the scenarios collected in the database will be used. This will permit the estimation of droplet and aerosol concentrations and distributions in unknown scenarios at low-computational cost, in near real-time. A state-of-the-art AI-based framework, incorporating descriptive, predictive and prescriptive techniques will extract the knowledge from the data and drive the decision-making process and provide in near real-time the assessment of risk levels.

One in 17 people have a rare disease. Rare diseases can be extremely difficult to diagnose, but they often have an unidentified genetic cause. Recent advances in clinical imaging, pathology, and genomic technologies have led to remarkable progress in understanding disease - particularly rare diseases. However, the power of these technologies cannot be fully realised until the immense volume of data generated can be integrated with NHS data, then analysed by researchers in a secure environment that protects the privacy of individuals. Working across the NHS, academia and industry we will use existing tools to transfer data from NHS Trusts to a secure environment that interfaces with the NHS network and shares data with Public Health England. NHS information will then be combined with research data in a cloud-based platform. Initially, we will involve patients with rare diseases recruited to the NIHR BioResource; a national resource of volunteers who have already provided consent that information retrieved from their health records can be used for medical research. This will create a rich research resource with the potential to transform our understanding of rare genetic disorders, drive improvements in diagnosis and management, and provide proof of principle for use in other diseases.

Antimicrobial resistance is one of the greatest public health threats spanning the One Health continuum (humans, animals and the environment). Antibiotics are of invaluable public health importance and are used on a daily basis worldwide to save and ease the suffering of millions of human and animal lives. However, their extensive and often uncontrolled use has led to the global spread of resistance in bacteria of medical and veterinary importance to an unprecedented level. This is threatening the ways we practice medicine and our ability to care for the sickest patients including those in need of life-saving treatments such as organ transplantation or cancer chemotherapy, and those in intensive care units. Antibiotic resistance is now recognised by the WHO as one of the greatest threats to human health and is increasingly topical within medical, veterinary and lay organisations of national and global reach. Enterococcus faecium, a bacterium carried harmlessly in the gut of humans and animals, has emerged as a leading cause of infections in critically ill and severely immunocompromised patients in hospitals. It has a propensity to accumulate and disseminate multiple antibiotic resistance determinants. Our previous work using a bacterial DNA fingerprinting technique called short-read whole-genome sequencing (WGS) established that E. faecium causing infections in hospital belongs to distinct strains from those found in livestock. In addition, we found different types of antibiotic resistance genes predominating in the two reservoirs. However, we also found instances of identical resistance genes, including to classes of antibiotics that are important in human medicine. Short-read WGS has limitations when trying to reconstruct the hierarchical levels of transmission units responsible for the spread of antibiotic resistance, which range from the whole bacterial strains, to consecutively smaller layers of mobile genetic elements known as plasmids and transposons down to the gene level. In order to decipher this "Russian doll" model, a different technique known as long-read WGS is required. Here, we propose to carefully select isolates for long-read WGS to allow us to quantify and understand the architectural context of shared antibiotic resistance genes between human and animal strains of E. faecium. Antibiotic susceptibility testing is a technique used daily in laboratories around the world to establish if antibiotics are still effective at treating bacterial strains of interest (i.e. ensuring the strains have not developed resistance). Resistance to antibiotics is mediated by genetic changes, hence whole genome sequencing has emerged as an attractive technology to characterise the full repertoire of known genetic changes that cause resistance and predict from the bacterial DNA if antibiotics are still effective. However, a complete understanding of the genetics governing resistance to antibiotics is required before WGS can be adopted to inform antibiotic prescribing. Our previous research has shown that WGS is very good at predicting the effectiveness of most antibiotics in E.faecium, except for 3 last-resort antibiotics used against the most resistant strains: daptomycin, tigecycline and linezolid. Here, we aim to redress this shortcoming by generating additional laboratory tests and sequencing data and to apply state-of-the art population genomic methods to improve predictions.