ASPIRE: Automated Sensing & Predictive Inference for Respiratory Exacerbation
Lead Research Organisation:
University of Oxford
Department Name: Engineering Science
Abstract
There is an urgent, unmet need for reliable, intelligent systems that can monitor patient condition in the home, and which can help patients manage long-term conditions. Delays in recognition of the changes in physiological state worsen outcomes and increase healthcare costs. The ASPIRE programme uses chronic obstructive pulmonary disorder (COPD) as an exemplar, which affects over 210 million people globally. This condition costs the National Health Service over £800 million each year, over half of which is spent treating patients in hospital, rather than caring for them in their homes.
Intelligent monitoring systems are required to address the needs of patients with long-term conditions in their homes. However, no wearable systems have penetrated into clinical practice at scale, due to: (i) poor tolerance of existing wearable devices for monitoring; (ii) a lack of robustness in the estimates of the vital signs that wearable sensors produce; (iii) very limited battery life that requires batteries to be re-charged at a rate that prevents their use on a large scale; and (iv) limited subsequent use of the data for helping the patient understand and manage their condition.
We propose to develop an "intelligent" home-based system, with smart algorithms embedded within lightweight healthcare sensors, to overcome these limitations. Our novel work will incorporate next-generation machine learning algorithms to combine information from healthcare sensors with information from GP and hospital visits. This will enable the system to learn "normal" health condition for individual patients, with knowledge of other conditions from which they may be suffering, and which can then make recommendations to the patient concerning self-management of their condition. This work will include close working with world-leading clinicians to ensure that the recommendations provided by the system are correct for the individual patient.
Intelligent monitoring systems are required to address the needs of patients with long-term conditions in their homes. However, no wearable systems have penetrated into clinical practice at scale, due to: (i) poor tolerance of existing wearable devices for monitoring; (ii) a lack of robustness in the estimates of the vital signs that wearable sensors produce; (iii) very limited battery life that requires batteries to be re-charged at a rate that prevents their use on a large scale; and (iv) limited subsequent use of the data for helping the patient understand and manage their condition.
We propose to develop an "intelligent" home-based system, with smart algorithms embedded within lightweight healthcare sensors, to overcome these limitations. Our novel work will incorporate next-generation machine learning algorithms to combine information from healthcare sensors with information from GP and hospital visits. This will enable the system to learn "normal" health condition for individual patients, with knowledge of other conditions from which they may be suffering, and which can then make recommendations to the patient concerning self-management of their condition. This work will include close working with world-leading clinicians to ensure that the recommendations provided by the system are correct for the individual patient.
Planned Impact
The proposed programme has the potential for very significant impact for patients in the home who are suffering from long-term conditions, by optimising self-management and improving outcomes. Patients who are deteriorating will be identified early, which will allow preventative action to be taken, avoiding serious escalation (such as unplanned presentation to a hospital Emergency Department), and which will reduce the incidence of preventable morbidity and death. Patient-specific care will be enabled using an intelligent system that learns patient characteristics in real-time, thus improving patient outcomes by improving the efficacy of care provided, and the recommendations made to patients to assist in (i) understanding their own condition and (ii) taking steps to maintain stability.
The translation of such systems into the home-care environment will provide benefit to patients with long-term conditions, such as chronic obstructive pulmonary disease (COPD, our exemplar) - this will be achieved via predictive systems that allow patients to track view their condition, and robust forecasts of their health status, without the false-alarm rate associated with existing systems, and which prevents existing systems providing benefit to existing patients. Healthcare workers, including community nurses and GPs, will benefit from the outputs of the programming by being given a quantitative assessment of cohorts of their home-based patients, and records of both their physiological data and associated predictions / recommendations made by the ASPIRE system.
The NHS as an organisation will benefit from the research because improved patient outcomes are associated with lower healthcare costs, as a result of fewer stays in hospital and fewer unplanned admission to (expensive) higher levels of care. Additionally, patients can be stratified according to risk of severity / deterioration, allowing improved use of community staffing resources. Clinicians will benefit by being able to interpret, for the first time, the very large and heterogeneous datasets that are available for their patients - enabled by robust, probabilistic tools created during the programme.
Companies in the commercial private sector will benefit from the research, where involvement of the industrial partners will allow rapid implementation of the techniques developed during the programme. Such companies have an interest in "intelligent healthcare" algorithms that can be integrated into existing healthcare IT products, which will add significant value and market differentiation. Additionally, the rapidly-growing market for wearable devices is currently focused only on consumers - the proposed work will extend this market to healthcare technologies, by exploiting the opportunities for large-scale innovation and clinical validation that exist in the programme. The UK economy will benefit by the possible creation of new spin-out activity based around the activity of the proposed work, in addition to the "first to market" advantage conferred on the industrial partners.
The scientific community, and the UK research base in particular, will benefit from developing capacity in an emerging field of global importance, and where the proposed project will train the next generation of researchers in intelligent community care, and integrated home/hospital provision. The methodology developed within the proposed programme will be of translational benefit to other scientific disciplines, including other computational and mathematical sciences. Results from the research will feed into the Alan Turing Institute, where researchers involved across the UK will benefit from the development of large-scale sensing methodologies for data science.
The public will benefit via public-engagement activities run in collaboration with charities described in "Pathways to Impact".
The translation of such systems into the home-care environment will provide benefit to patients with long-term conditions, such as chronic obstructive pulmonary disease (COPD, our exemplar) - this will be achieved via predictive systems that allow patients to track view their condition, and robust forecasts of their health status, without the false-alarm rate associated with existing systems, and which prevents existing systems providing benefit to existing patients. Healthcare workers, including community nurses and GPs, will benefit from the outputs of the programming by being given a quantitative assessment of cohorts of their home-based patients, and records of both their physiological data and associated predictions / recommendations made by the ASPIRE system.
The NHS as an organisation will benefit from the research because improved patient outcomes are associated with lower healthcare costs, as a result of fewer stays in hospital and fewer unplanned admission to (expensive) higher levels of care. Additionally, patients can be stratified according to risk of severity / deterioration, allowing improved use of community staffing resources. Clinicians will benefit by being able to interpret, for the first time, the very large and heterogeneous datasets that are available for their patients - enabled by robust, probabilistic tools created during the programme.
Companies in the commercial private sector will benefit from the research, where involvement of the industrial partners will allow rapid implementation of the techniques developed during the programme. Such companies have an interest in "intelligent healthcare" algorithms that can be integrated into existing healthcare IT products, which will add significant value and market differentiation. Additionally, the rapidly-growing market for wearable devices is currently focused only on consumers - the proposed work will extend this market to healthcare technologies, by exploiting the opportunities for large-scale innovation and clinical validation that exist in the programme. The UK economy will benefit by the possible creation of new spin-out activity based around the activity of the proposed work, in addition to the "first to market" advantage conferred on the industrial partners.
The scientific community, and the UK research base in particular, will benefit from developing capacity in an emerging field of global importance, and where the proposed project will train the next generation of researchers in intelligent community care, and integrated home/hospital provision. The methodology developed within the proposed programme will be of translational benefit to other scientific disciplines, including other computational and mathematical sciences. Results from the research will feed into the Alan Turing Institute, where researchers involved across the UK will benefit from the development of large-scale sensing methodologies for data science.
The public will benefit via public-engagement activities run in collaboration with charities described in "Pathways to Impact".
Publications
Zhu T
(2021)
Comparison of parametric and non-parametric Bayesian inference for fusing sensory estimates in physiological time-series analysis.
in Healthcare technology letters
Zhu T
(2019)
Unsupervised Bayesian Inference to Fuse Biosignal Sensory Estimates for Personalizing Care
in IEEE Journal of Biomedical and Health Informatics
Zhu T
(2018)
Bayesian fusion of physiological measurements using a signal quality extension.
in Physiological measurement
Yang Y
(2018)
Machine learning for classifying tuberculosis drug-resistance from DNA sequencing data.
in Bioinformatics (Oxford, England)
Yang J
(2022)
Machine Learning-Based Risk Stratification for Gestational Diabetes Management.
in Sensors (Basel, Switzerland)
Velardo C
(2021)
Toward a Multivariate Prediction Model of Pharmacological Treatment for Women With Gestational Diabetes Mellitus: Algorithm Development and Validation.
in Journal of medical Internet research
Vasey B
(2022)
Reporting guideline for the early stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI.
in BMJ (Clinical research ed.)
Description | We have investigated improving the state-of-the-art with sensor inference, and developed deep learning models that perform well in new datasets and collaborations. |
Exploitation Route | Patents, publications. |
Sectors | Healthcare |
Description | We have conducted very large studies of wearables for use in clinical practice, and are in the process of specifying and managing clinical data-collection studies. I was awarded two national awards, leading to a new Chair at the University of Oxford. Patents arising from this work are in the process of being registered, with the potential for a new spin-out to be created after the lifetime of the award. Substantial funding applications are in review from several research councils; the work led to the establishment of a £30m centre funded by the Innovation & Technology Commission of Hong Kong. |
First Year Of Impact | 2019 |
Sector | Digital/Communication/Information Technologies (including Software),Healthcare |
Impact Types | Societal Economic |
Description | Healthcare Wearables for Independent Living |
Amount | £1,216,069 (GBP) |
Funding ID | EP/W031744/1 |
Organisation | Engineering and Physical Sciences Research Council (EPSRC) |
Sector | Public |
Country | United Kingdom |
Start | 01/2023 |
End | 12/2025 |
Description | Research Chair |
Amount | £1,586,000 (GBP) |
Organisation | Royal Academy of Engineering |
Sector | Charity/Non Profit |
Country | United Kingdom |
Start | 04/2023 |
End | 04/2028 |
Description | Research Professorship |
Amount | £1,823,387 (GBP) |
Funding ID | NIHR302440 |
Organisation | National Institute for Health Research |
Sector | Public |
Country | United Kingdom |
Start | 12/2022 |
End | 11/2027 |
Description | Chinese University of Hong Kong |
Organisation | Chinese University of Hong Kong |
Country | Hong Kong |
Sector | Academic/University |
PI Contribution | This partnership involved the founding of the Centre for Cardiovascular Engineering, funded by £31m from the Innovation & Technology Commission of Hong Kong. |
Collaborator Contribution | Wearable sensing, AI for wearable sensor data, healthcare technologies. |
Impact | Patents, publications. |
Start Year | 2023 |
Description | Research collaboration with GlaxoSmithKline |
Organisation | GlaxoSmithKline (GSK) |
Country | Global |
Sector | Private |
PI Contribution | Jointly working on AI methods for improving healthcare and the development of medicines. |
Collaborator Contribution | Data, domain expertise, problem setting. |
Impact | Open-access publications, support for national awards. This is a multi-disciplinary collaboration between AI scientists, clinicians, medical statisticians, and medical scientists. |
Start Year | 2021 |
Title | DEEP END-TO-END CLASSIFICATION OF ELECTROCARDIOGRAM DATA |
Description | There is disclosed a computer-implemented method of classifying electrocardiogram data of a patient, comprising the steps of receiving input data from each of a plurality of electrocardiogram leads, arranging the input data into a single combined image, and applying a machine-learning algorithm to the combined image to classify the electrocardiogram data. |
IP Reference | WO2021014150 |
Protection | Patent application published |
Year Protection Granted | 2021 |
Licensed | Commercial In Confidence |
Impact | Further development via industry. |
Title | METHOD AND APPARATUS FOR MONITORING A HUMAN OR ANIMAL SUBJECT |
Description | Methods and apparatus for monitoring a human or animal subject are disclosed. In one arrangement, measurement data representing a time series of measurements on a subject is received. The measurement data is represented as a mathematical expansion comprising a plurality of expansion components and expansion coefficients. First and second partial reconstructions are performed using first and second subsets of the expansion components. First and second spectral analyses are performed on the first and second partial reconstructions to determine first and second dominant frequencies. A frequency of a periodic physiological process is derived based on either or both of the first and second dominant frequencies. |
IP Reference | US2021000384 |
Protection | Patent application published |
Year Protection Granted | 2021 |
Licensed | Yes |
Impact | Licensed to Sensyne Health Plc for translation into products. |
Title | METHOD AND APPARATUS FOR MONITORING A HUMAN OR ANIMAL SUBJECT |
Description | Methods and apparatus for monitoring a human or animal subject are disclosed. In one arrangement, test data representing a time-series of physiological measurements performed on a subject in a measurement session is received. A mean trajectory with error bounds is fitted to the test data. A state of the subject is determined by comparing the fitted mean trajectory with error bounds to a stored model of normality. The stored model of normality comprises a library of latent mean trajectories with error bounds. Each latent mean trajectory with error bounds is derived by fitting a hierarchical probabilistic model to a respective one of a plurality of sets of historical data. Each set of historical data comprises a plurality of session data units. Each session data unit representing a time-series of physiological measurements obtained during a different measurement session. The latent mean trajectory with error bounds for the set describes an underlying function governing each of the time-series of the session data units of the set. |
IP Reference | US2020395125 |
Protection | Patent application published |
Year Protection Granted | 2020 |
Licensed | Commercial In Confidence |
Impact | Further development via industry. |