Artificial intelligence programmed to have “imagination” could help doctors diagnose illnesses, new research suggests.
A study found that AI trained to be able to use causal reasoning can be just as accurate as human doctors in diagnosing patients.
Researchers at healthcare technology firm Babylon used an approach known as causal machine learning, where AI is given “imagination” to consider what symptoms it might see if the patient had a different illness from the one it was considering.
Dr Jonathan Richens, lead author of the study and a Babylon scientist, said: “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?’
“This allows the AI to tease apart the potential causes of a patient’s illness and score more highly than over 70% of the doctors on these written test cases.”
The research, which has been published in Nature Communications, saw more than 1,600 realistic written medical cases created by Babylon GPs, which included typical and atypical examples of symptoms.
Another group of GPs, as well as an older AI algorithm and the new causal machine learning AI, were then asked to name the illnesses they considered most likely for each of the cases.
According to the study, the new AI performed best, naming the correct illness in its answers on average 77% of the time, ahead of the older AI (72.5%) and the human doctors (71.4%).
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 30% of both rare and very rare conditions.”
Babylon said it hoped the technology could be used in the future to help speed up diagnosis, improve accuracy and free up time for clinicians.
The health tech firm said the algorithm would face further development and testing before it would be considered for release in the company’s public app.