Industry researchers in medical AI have demonstrated a “causal reasoning” algorithm whose diagnostic accuracy would place it among the top 25% of primary care doctors in the U.K.
By comparison, a standard associative algorithm landed among the top 48% of the 44 physicians in the study cohort.
Associative algorithms use linear deductions to match symptoms with diseases likely to cause them. Causal reasoning AI—also known as counterfactual AI or “AI with imagination”—considers whether a different disease might be causing a given symptom or set of symptoms.
The researchers, three present or former executives with the London-based digital health company Babylon Health, describe the work in a study published by Nature Communications.
In the study’s faceoffs on overall diagnostic accuracy, the physicians had a mean score of 71.4% vs. 72.5% for the older algorithm, which was created for the research.
Both were topped by the new causal algorithm, which scored 77.3%. This was higher than 32 of the doctors, equal to one and lower than 11.
Further, the new algorithm was especially impressive at diagnosing rare and very rare conditions.
“While we have focused on comparing our algorithms to doctors, future experiments could determine the effectiveness of these algorithms as clinical support systems—guiding doctors by providing a second opinion diagnosis,” the authors write in their discussion section. “Given that our algorithm appears to be complimentary to human doctors, performing better on vignettes that doctors struggle to diagnose, it is likely that the combined diagnosis of doctor and algorithm will be more accurate than either alone.”
In a news release, Babylon’s founder and CEO, Dr. Ali Parsa, places the achievement in the context of poor healthcare access prevalent in much of the world.
The experimental algorithm’s success in the study “should not be sensationalized as machines replacing doctors, because what is truly encouraging here is for us to finally get tools that allow us to increase the reach and productivity of our existing healthcare systems,” Parsa says. “AI will be an important tool to help us all end the injustice in the uneven distribution of healthcare, and to make it more accessible and affordable for every person on Earth.”
The journal has posted the study in full for free.