AI database used to improve treatment of UK COVID-19 patients
NHSX has pioneered an anonymous bank of over 40,000 scans collected over the course of the pandemic. The National COVID-19 Chest Imaging Database (NCCID) collects X-Rays, CT and MRI scans to speed up diagnosis of the disease, enabling clinicians to implement treatment plans faster and more effectively as well as allowing them to predict more accurately which patients may end up in a critical condition.
Established in April 2020 with the British Society of Thoracic Imaging (BSTI), Royal Surrey NHS Foundation Trust and AI specialists Faculty, the NCCID provides a more comprehensive understanding of COVID-19, collating data from 20 NHS trusts across the country to detect disease patterns and markers of the virus, which, in turn, enable more accurate early diagnoses and prognoses and inform new treatments. This information, which is stripped of any patient identifying information, is also used to inform algorithms used in hospitals and universities to treat and monitor patients.
A particular benefit highlighted was more accurate early detection. For example, clinicians at Addenbrooke’s Hospital in Cambridge are using the information gathered on the database to build an algorithm to better identify whether a patient exhibiting potential early COVID-19 symptoms has the virus even before they have received a positive test. This is achieved by matching visual signatures of the virus held in the NCCID with incoming patients’ imaging results.
By understanding the earlier stages of the disease and enabling faster diagnosis, clinicians are able to anticipate potential complications and provide earlier medical interventions to stop them. This may mean giving patients medication or oxygen before they reach critical condition or predicting ICU capacity and managing staffing accordingly.
WHY IT MATTERS
The NCCID enables medical professionals to grasp the true extent of COVID-19 in the UK. With images stored from over 10,000 patients so far, it highlights the variations and different manifestations of the virus and avoids data biases of, for example, patient geography, ethnic background or gender, which may arise in more isolated attempts at data collection. The availability of the NCCID also ensures a safer, more consistent therapeutic response to COVID-19 across the country.
The breadth of the NCCID is proving hugely helpful in research. Carola-Bibiane Schönlieb, head of the Cambridge Image Analysis group at the University of Cambridge, called the resource “invaluable” as it “provided [them] with a diverse, well-curated, dataset of UK patients to use in our algorithm development.”
She continued: “The ability to access the data for 18 different trusts centrally has increased our efficiency and ensures we can focus most of our time on designing and implementing the algorithms for use in the clinic for the benefit of patients.”
The NCCID also assists in the development and training of AI tools by companies to assist in the COVID-19 effort. The NCCID provides a large, representative dataset with which to test and validate these models, particularly those born out of chest imaging data, ensuring that these tools are developed to be as safe and effective as possible.
THE LARGER PICTURE
In the longer term, the NCCID is ensuring the future of data sharing in the NHS. Jennifer Berger, AI Engagement Lead at NHSX, told HealthcareITNews: “A nationally co-ordinated approach [such as the NCCID’s] ensures that data sharing agreements are put in place at speed whilst meeting the full requirements of data protection legislation and the highest standards of Information Governance.”
By pioneering this national collaboration, the NCCID is helping pave the way for future AI imaging platforms for other health conditions such as heart disease and cancers and ensuring preparedness for future pandemics.
Accordingly, NHSX has invited all trusts with a radiology department to join the database so as to guarantee increasingly representative data in the NCCID and to further develop and test the performance of the AI model.
Praise of the NCCID comes alongside the release of a new NHS publication. A guide to good practice for digital and data-driven health technologies, published this week, outlines what the NHS looks for when purchasing new digital technologies for health and care.
ON THE RECORD
Dominic Cushnan, head of AI Imaging at NHSX, praised the NCCID, saying: “There is huge potential for patient care, whether through quicker analysis of chest images or better identification of abnormalities. The industrial scale collaboration of the NHS, research and innovators on this project alone has demonstrated the huge potential and benefits of technology in transforming care.”
Matt Hancock, Secretary of State for Health and Social Care, added: “The use of artificial intelligence is already beginning to transform patient care by making the NHS a more predictive, preventive and personalised health and care service. It is vital we always search for new ways to improve care, especially as we fight the pandemic with the recovery beyond. This excellent work is testament to how technology can help to save lives in the UK.”