Advertisement

Babylon reveals AI with ‘imagination’ to help doctors better diagnose disease

CEO of Babylon, Ali Parsa - Geoff Pugh  /Geoff Pugh  
CEO of Babylon, Ali Parsa - Geoff Pugh /Geoff Pugh

Digital doctor app Babylon Health has revealed an artificial intelligence system with “imagination” that it claims can more accurately assess patient symptoms and support doctors on the frontline.

A study, published in peer-reviewed journal Nature Communications, showed that an AI programmed by Babylon researchers to use “causal reasoning” could offer a more accurate diagnosis of a patient versus existing algorithms.

Doctors assessing patients must consider a range of different diagnoses based on the symptoms they are presented with, a task that requires them to “imagine” what a patient has been affected by.

Existing AI systems, which learn to spot correlations and patterns in data, are unable to do this.

The new algorithm allows the AI to consider what symptoms “it might see” if the patient had a different illness from the one it was considering, a key factor in doctors’ assessments of patients.

“We took an AI with a powerful algorithm, and gave it the ability to imagine alternate realities and consider ‘would this symptom be present if it was a different disease?’” said Dr Jonathan Richens, a scientist at Babylon and lead author of the research. “This allows the AI to tease apart the potential causes of a patient’s illness.”

Babylon, backed by Health Secretary Matt Hancock, currently operates AI in a chatbot available to patients in London, Birmingham, Manchester and Wolverhampton.

To determine the accuracy of the new AI, Babylon created over 1,671 different medical cases that included typical atypical symptoms of more than 350 illnesses, though it has faced some criticism over its accuracy.

Read More | Babylon's NHS Kingdom
Read More | Babylon's NHS Kingdom

The cases were then assessed by a group of 44 Babylon GPs, an old algorithm and the new AI system, all of which listed the illnesses they considered most likely based on the information at hand.

According to the study, the new AI performed best, naming the correct illness in its answers on average 77pc of the time, ahead of the older AI (72.5pc) and the human doctors (71.4pc).

Dr Saurabh Johri, chief scientist at Babylon, said the results suggested human doctors and AI could work together in the future.

"Interestingly, we found that the AI and doctors complemented each other - the AI scored more highly than the doctors on the harder cases, and vice versa," he said.

"Also, the algorithm performed particularly well for rare diseases which are more commonly misdiagnosed, and more often serious. Switching from using correlations improved accuracy for around 30pc of both rare and very rare conditions."

Dr Richens added: “We observed that the cases that the humans struggled with, the AI did very well in and vice versa, so for some reason these two diagnostic agents complement each other and do better together,” he said.

The new algorithm is not yet publicly available on Babylon’s app, and will only be released after further development and once regulatory approvals in the UK have been attained.

Would you use a digital doctor app? Tell us in the comments section below