Commits to supporting government plans in partnership
with leading UK academics and NHS Trusts to significantly reduce deaths from
chronic diseases using its clinical artificial intelligence tools and expertise
- Drayson Health has been working with the Oxford University Hospitals NHS Foundation Trust and University of Oxford in a unique collaboration in the clinical artificial intelligence space developing clinician-led software applications that improve patient care and reduce healthcare costs
- Working alongside medical researchers from Oxford, using our AI software tools and expertise, we are unlocking clinical evidence from real-word patient data conforming to the highest ethical and patient privacy standards to deliver AI powered regulated digital tools that improve health, advance research programmes, accelerate the discovery
of new therapeutic treatments and uncover new indications for existing drugs
- Drayson Health’s unique partnership with Oxford was formed in July 2017 underpinned by a Strategic Research Agreement that will ensure new technologies and digital innovations born out of clinical need can be safely commercialised and scaled to benefit more NHS organisations and the patients they serve, and exported internationally.
- Drayson Health’s unique approach also means a commitment to ‘giving back’ to the NHS and Higher Education sector through a royalty and equity ownership in our business.
- Our technologies and research support programmes in chronic diseases such as diabetes, cardiovascular conditions and respiratory disease as well as cancer and Parkinson’s disease.
- We have already seen the benefits of our model through wider adoption of the software applications across the NHS (SENDTM, GDm-HealthTM).
- Drayson Health is headquartered at the Big Data Institute (BDI) at the University of Oxford’s Old Road Campus. Former Science Minster Lord Drayson leads the Drayson Health team, which combines experience in AI, digital health and life sciences. The team is advised by an eminent Scientific Advisory Board, chaired by Regius Professor of medicine, Sir John Bell
Professor Lionel Tarassenko CBE FREng FMedSci, Head of Engineering at Oxford, said:
very much welcome the Prime Minister’s speech announcing the Government’s
ambitious plans to transform outcomes for people with chronic diseases. For the
past decade, the University of Oxford and the Oxford University Hospitals NHS
Foundation Trust have been working in partnership to apply the power of
Artificial Intelligence, especially machine learning, to healthcare data in
order to improve patient outcomes. This world-leading partnership between
engineers, computer scientists and clinical researchers has been funded by NIHR
in the Oxford Biomedical Research Centre, the Department of Health and the Wellcome
Trust and has led to the creation of the Big Data Institute on the medical
campus in Oxford.
have designed AI-powered digital tools for patients with chronic diseases such
as diabetes or COPD to improve their self-management and self-care. We have
assembled highly-curated databases of patient-generated data, which we can now
link to in-hospital data. By applying the latest machine learning techniques
developed in the University to these NHS datasets, we can begin to deliver on
the ambition of helping people enjoy an additional five years
of healthy, independent life. We already have promising results
from our machine learning techniques showing how the risk of a major,
debilitating stroke can be reduced by better targeting of medicines for those
individuals at risk. In the past year, the University and the NHS Trust have
signed a Strategic Research Agreement with Drayson Health to enable our digital
tools and data analytics powered by machine learning to be scaled up beyond
Oxford, and made available to the whole of the NHS.”
The process behind Drayson
- Patients and clinicians use our medical software, which is clinically proven to improve outcomes.
- Our software collects data on the care of patients in the real world which is then anonymised.
- We combine this anonymised patient data with other data to create phenotypic datasets of value for discovery research.
- We use A.I. to analyse the data & answer clinically relevant questions and generate new discovery hypotheses.
- This enables improvements in pharmaceutical development and clinical trials...
- ... and the discovery of new medicines that further improve patient outcomes.
The history of Drayson
Health and Clinical AI
A branch of research into artificial intelligence at Oxford
and upon which our work at Drayson Health is based is becoming commonly known
as CLINICAL AI. This is a particular
field of artificial intelligence concerned with the analysis of data acquired
during routine clinical care of patients. Clinical AI is a distinct subset of
AI research comprising approaches that have been developed based upon the prior
knowledge of expert clinicians encoded in the models themselves.
Clinical data is “messy” - it has missing elements, is
noisy, often contradictory, dense, may include errors in recording, and many
other artefacts. The variability between humans is such that conventional,
generic “one-size-fits-all” models of patient health do not lead to improving
our understanding of the data; additionally, the data are typically very
“deep”, comprising recordings of very many different types (vital signs,
blood-test results, medications, etc.) all taken at different times, and under
different conditions. These are
especially difficult to understand and model without extremely close
“co-design” of AI systems with medical doctors.
This analysis of so-called “longitudinal” data for a patient
over time often takes into account THOUSANDS of variables related to the health
of a patient. The human body exhibits homeostasis (i.e., it seeks to return to
normal following a disturbance to the system such as that caused by a disease),
and many of these variables are connected. For example, blood pressure,
breathing rate, heart rate, and the amount of oxygen in the blood (the “vital
signs”) change in a way that varies depending upon the individual, and with the
severity and type of illness. Developing AI models that are sufficiently robust
for use in clinical practice, and which are sufficiently reliable to inform how
patients should be treated, is a computing challenge of huge complexity and
requires DEEP CLINICAL KNOWLEDGE of how the body works and how healthcare is
delivered. The purpose of clinical AI is to be able to develop
computer models that are
useful in analysing the data effectively in the real world and that are
sufficiently understandable and predictable to be used in a highly regulated
environment such as healthcare.
The other important barrier to building expertise in
Clinical AI is the question of data governance. The regulatory requirements that
people must meet to be allowed to have access to clinical data on patients are
considerable – for obvious reasons. This is highly personal, private data. It
has to be properly consented and must be treated with the utmost care and
security. That is why most clinical AI groups only work on publicly available
data that has already been anonymised. Even the process of anonymising data is
complex and requires considerable expertise – of both clinical practice and
Oxford has been a pioneer in this area of health
informatics, neural networks and medical sensor data analysis since the early
2000s when work by Lionel Tarassenko at the Medical Engineering Unit (where
PowderJect Pharmaceuticals started) began on the development of sensor devices
for measuring vital signs and the development of signal processing algorithms
to ensure the validity and quality of the measurement data they provided.
This work received a massive boost in 2008 with the creation
of the Institute for Biomedical Engineering in part funded by PowderJect
co-founder Prof Brian Bellhouse, based on the vision of Prof John Bell to have
bio-medical engineers co-located with the medics in the nearby hospitals. Then
the work took off, boosted by winning a series of NIHR research grants focused
on the use of machine learning and signal processing techniques to monitor
patients remotely and to collect high quality, well curated data on patients
and their treatment in the real world via a series of substantive clinical trials.
The development of the SEND system led by Prof Peter
Watkinson beginning in 2012 was the next major milestone and as SEND was
adopted throughout the hospitals of the OUH during 2015 and 2016 the amount of
time series data being collected on patients saw a massive jump.
At that time an application was successful for research
funding to create a broader, well curated database that included multiple sets
of patient data such as prescriptions, diagnoses etc as well as vital signs
data. This was aimed at developing a risk index for individual patients, that
would be based on an analysis of their individual time-series dataset to
ascribe a risk of deterioration, helping doctors to identify patients of risk
of deterioration, before it is too late for an intervention to reverse the
deterioration. In 2014 the grant from the Department of Health and the Wellcome
Trust to support the HAVEN project was another important step and led to the
creation of a very large database of well curated patient data at Oxford.
As the database increased in size and became massive it
required the development of new tools to enable sense to be made of the
complexity of the data sets and this led to the creation of the research theme
in Oxford of Clinical AI, led by Profs Lionel Tarassenko, Peter Watkinson and
David Clifton, and a world leadership position in the field.
Following the effective use of these tools in Oxford it was
clear that the opportunity to scale these systems so that they could be
deployed more widely across the NHS and potentially exported internationally
required very significant investment and commercial coding skills and resources
beyond the remit of a research University and a NHS Trust. In 2017 a Strategic
Research Agreement was established with Drayson Technologies to bring these
additional resources to bear and to provide a structure whereby the commercial
returns from the wider use of these innovations could be re-invested back in
the Trust and the University via an equity and royalty agreement.