Developing valuable use cases to accelerate the adoption of AI
We designed and implemented the vision for a new AI lab to support safe and effective deployment of AI technologies.
Background
We designed and implemented the vision for a new AI lab to support safe and effective deployment of AI technologies.
“AI imaging technologies are experiencing rapid advances, but the real challenge is how you deploy them into Trusts and ensure they perform as expected. The NCCID and validation programme were instrumental in understanding this better.”
Rosalind Berka, Senior Manager
Faculty
Solution
Along with the NHSX and a consortium of academic research groups, including Royal Surrey NHS Foundation and the British Society of Thoracic Imaging, we designed and put to the test two new case studies to showcase the AI lab’s capabilities. One created the UK’s first centralised chest imaging database to help support the development of third-party tools, which AI developers can use to better train their models and improve performance. The second was designing and implementing a validation process for machine-learning models which takes companies through a rigorous monitoring phase to build up evidence for regulatory approvals and ensures that models are safe and effective for use on patients.
Impact
Along with the NHSX and a consortium of academic research groups, including Royal Surrey NHS Foundation and the British Society of Thoracic Imaging, we designed and put to the test two new case studies to showcase the AI lab’s capabilities. One created the UK’s first centralised chest imaging database to help support the development of third-party tools, which AI developers can use to better train their models and improve performance. The second was designing and implementing a validation process for machine-learning models which takes companies through a rigorous monitoring phase to build up evidence for regulatory approvals and ensures that models are safe and effective for use on patients.
“This external validation process is critical, especially given the lack of model validation of AI models used in radiology. Faculty and the consortium have opened up vast opportunities not just for future AI adoption in the NHS, but for developers to build better performing and safer AI models, too.”