6 days ago
NHS Lothian partner in project evaluating adoption of artificial intelligence products
Testing a single piece of AI software currently takes more than nine months
NHS Lothian is one of two health boards in Scotland partnering with an evaluation company to speed up the safe adoption of artificial intelligence (AI) in the healthcare system.
The £1 million Innovate UK-funded project brings together NHS Lothian and NHS Greater Glasgow and Clyde hospitals, Edinburgh and Glasgow universities, and technology companies.
Testing a single piece of AI software currently takes more than nine months, and with more than 200 AI products available for a single hospital department, checking all of them individually using traditional clinical trial or service evaluation models isn't realistic. The result is that very few trials turn into real-world solutions, meaning that patients, hospitals and AI developers all lose out.
The collaboration will use Aival's independent evaluation platform to assess AI systems in a secure environment, with all patient data anonymised and remaining within an NHS secure environment.
The project will assess the use of the Aival platform to compare six commercial AI products for the triage of stroke and lung cancer patients from AI developers InferVision, Annalise-AI and – who are all collaborating with the programme.
AI products are often calibrated to the locations where they are being used to allow for differences in the local population, meaning they may not generalise to different groups without extra training.
The performance of AI products changes over time – even when the product itself isn't altered. This is known as 'drift', and can happen as a patient population changes, as the commonness of a disease shifts, or a scanner's software is updated.
Dr Rishi Ramaesh, Consultant Radiologist and Innovation Fellow at NHS Lothian, said: 'Artificial intelligence has tremendous potential to improve patient care, but healthcare leaders need confidence that these systems are safe and effective. This project will help healthcare leaders to evaluate AI, and make sure that new technologies deliver real benefits for patients.'
Luciana D'Adderio, Edinburgh University academic and AI evaluation and assurance expert, said: 'AI technology is achieving widespread deployment across healthcare settings, yet its assurance has not received the rigorous attention it demands.
'The current frameworks for evaluation and monitoring fall short in adequately assuring AI systems, primarily because AI represents a fundamental departure from the traditional technologies these frameworks were designed to assess.
'There is an urgent need for innovative tools and technologies for AI assurance, which themselves must undergo thorough evaluation and validation. This critical challenge forms the cornerstone of our groundbreaking project.'
Kanwal Bhatia, CEO and founder of Aival, said: 'It's vital that we monitor and check AI that's being used in decisions on patients' health, to ensure the best outcomes for patients.
'Any healthcare workers using AI need to be sure that the product is doing what it says it does – not just now, but five years in the future. While many NHS teams understand this, they currently lack the tools to properly evaluate the AI products.
'Putting in place effective validation systems will encourage trust and adoption of AI, and will deliver cost savings and growth in the NHS and in private healthcare.
'We work hand in hand with NHS leaders, clinical and technical teams to provide the expertise and software to ensure that their AI systems are doing what they're supposed to do.'
