Scientists from University Hospitals Birmingham NHS Foundation Trust have found that artificial intelligence (AI) is on a par with human doctors in terms of its ability to accurately diagnose illnesses based on medical images.

Deep learning enables the AI to identify patterns of disease by examining thousands of medical images.

The researchers claim the technology has enormous potential to benefit the healthcare system.

The say it will ease strain on the NHS’ strained resources, free up time for doctor-patient interactions and aid the development of bespoke treatment.

Based on data gathered from 14 trials, the researchers found that deep learning correctly detected disease in 87% of cases – compared to 86% achieved by doctors.

In terms of being able to accurately eliminate healthy patients, doctors and AI were on a par, with the tech achieving 93% accuracy while doctors scored 91%.

Recommended

• Economy Secretary Pledges Support for Safer Business Stronger Scotland Campaign
• Which? Claims Booking.com is Still ‘Duping Consumers’
• Three New Appointments for Airts as Firm Scales to Meet Global Demand

“We found deep learning could indeed detect diseases ranging from cancers to eye diseases as accurately as health professionals, said lead author Professor Alastair Denniston.

“But it is important to note AI did not substantially out-perform human diagnosis.”

Dr Xiaoxuan Liu, the lead author of the study and from the same NHS trust, agreed. “There are a lot of headlines about AI outperforming humans, but our message is that it can at best be equivalent,” she said.

Denniston and his colleagues have cautioned that the findings are based on a small number of studies, due to the lack of high-quality research on the topic.

“Diagnosis of disease using deep learning algorithms holds enormous potential,” Denniston said.

“We reviewed over 20,500 articles, but less than 1% of these were sufficiently robust in their design and reporting that independent reviewers had high confidence in their claims.

“What’s more, only 25 studies validated the AI models externally – using medical images from a different population – and just 14 studies actually compared the performance of AI and health professionals using the same test sample.”

Denniston and his team are calling for higher standards of research and reporting to improve future evaluations.

“Evidence on how AI algorithms will change patient outcomes needs to come from comparisons with alternative diagnostic tests in randomised controlled trials,” said co-author Dr Livia Faes, of Moorfields Eye Hospital, London.