The National COVID-19 Chest Imaging Database (NCCID), a joint initiative established by NHSX, the British Society of Thoracic Imaging (BSTI), Royal Surrey NHS Foundation Trust and Faculty.
Chest scan images are a vital tool for researchers in the fight against COVID-19. However, siloed, incomplete datasets often make it difficult for researchers and developers to rigorously test hypotheses or tools that could help improve the healthcare response to the disease and save lives.
The NCCID is designed to provide researchers and technology developers with managed access to a complete, and representative sample of pseudonymised chest scans from suspected COVID-19 patients across the country. With this data, the NHS hopes to enable the development of software that can help doctors and researchers understand the impact and progression of the disease, assess the severity of the condition in individual patients, identify factors that may complicate recovery, and prioritise patients whose condition is most likely to deteriorate.
As part of our ongoing work on the NHSX AI Lab, Faculty was asked to provide the platform infrastructure and establish the processes required to ensure the success of the NCCID and enable secure access to the data.
We developed a secure, shared infrastructure – stored and owned by the NHS – that allows researchers and developers to safely access the data needed to develop and test technology solutions that target COVID-19. Crucially, we supported this infrastructure by establishing rigorous data management processes that ensure data is collected and accessed securely, efficiently, and safely.
As part of this process, we created a separate dataset that’s used solely to validate AI models and never for model training, protecting the integrity of the validation process. In doing so, the infrastructure is able to help organisations address a key barrier to AI implementation in healthcare: with no firm, AI-specific guidelines currently in place, healthcare providers often have little assurance that their AI purchase will deliver real, measurable results for practitioners and patients. Now, AI tools can be validated within the NCCID infrastructure platform to ensure that they are safe and effective for use in clinical settings.
Researchers or developers wishing to access the database and use the platform must have their requests assessed by a committee of scientific, technology, information governance and ethics experts. All of the scans in the library are stripped of any identifying patient details by each hospital trust before being submitted to the national collection, so that researchers only have access to pseudonymised information.
Dominic Cushnan, Head of AI Imaging at NHSX, said: “Applying the power of Artificial Intelligence to over 40,000 NHS scans is a fantastic opportunity to help us quickly detect disease patterns and treat patients with this new virus.”
The NCCID is now the biggest image library of its kind in the UK, containing around 60,000 scans from 20,000 patients in all, submitted by 22 organisations and NHS trusts. There are currently enough COVID positive images to build a validation set that can ensure AI tools currently in development meet the technical standards required for use in the NHS.
According to NHSX, teams from 11 British university consortiums are already using the NCCID to develop and test tools for the diagnosis, management and prioritisation of COVID-19 patients.