A recent study of Alphabet Inc’s DeepMind AI unit, which merged with Google Health last year, found that the system could detect cancers with a similar degree of accuracy to expert radiologists while also reducing false results.
With two different sample groups, one based in the UK and one in the US, the study showed that the number of false positives was reduced by 5.7% in the US-based group and by 1.2% in the British-based group. False negatives, where the tests were classified as normal, were cut by 9.4% in the U.S. group, and by 2.7% in the British group.
The AI was asked to look at and diagnose scans from around 29,000 women and come up with a diagnosis. The same scans were given to six radiologists, with the AI system outperforming all of them.
According to Reuters, while computers have been used to help with cancer detection for years, these computer-aided detection (CAD) systems are not as effective as the Google AI system because they are only trained to detect what radiologists can see.
By comparison, AI scans for cancers based on information collated from thousands of mammograms. This means they can “see” beyond limitations of the human eye and see subtle cues for cancer in the scans.
The idea behind improved detection would be to treat cancers as quickly as possible, however this study did not show improved outcomes for patient care.