The current tests miss 20 percent of breast cancers. While as reported by the Wall Street Journal the AI does better than doctors, it nevertheless misses some cancers that radiologists find: “Google’s health research unit said it has developed an artificial-intelligence system that can match or outperform radiologists at detecting breast cancer, according to new research. But doctors still beat the machines in some cases.”
Mammograms effective but have problems
Breast cancer is the second leading cause of cancer deaths in women, only lung cancer is more deadly and prevalent. Even though mammagrams are an effective common detection tool they nevertheless miss a large number of positive cases. Shravya Shetty a Google researcher and co-author of the paper said: “Mammograms are very effective but there’s still a significant problem with false negatives and false positives.”
The study
The Google-financed study used anonymized mammograms from more than 25,000 women from both the UK and another 3,000 from the US. Shetty said that researchers attempted to follow the same principles as those of radiologists. The research team first trained AI to scan X-ray imagines to look for signs of breast cancer by identifying changes to the breasts of the 28,000 women and they then checked the AI program’s analyses against that of the actual medical outcomes of the women.
The results
The researchers were ultimately able to reduce false negatives by 9.4 percent and false positives by 5.7 percent in the US. In the UK where two radiologists typically double-check the results the results were still positive but not by nearly the same degree. False negatives were cut by 2.7 percent but false positives by only 1.2 percent.
A recent Wired article notes: “The study claims the DeepMind algorithm performs better than a single radiologist, and is “non inferior” versus two.”The model performs better than an individual radiologist in both the UK and the US,” Kelly says. “In the UK we have this double reading system, where two radiologists or maybe three or four look at each scan… we’re statistically the same as that, but not better than that.” ” However, the article notes that the UK is facing a shortage of radiologists.
The AI program called Deep Mind was not perfect. In some cases radiologists flagged cases which the AI system missed even though Deep Mind outperformed examinations by single cardiologists,
Google hopes AI system can be used in clinical settings
Daniel Tse a Google product manager and co-author of the paper said the research team was working to make sure the findings could be generalized across populations: “We’re very excited and encouraged by these results. There’s obviously quite a bit of nuance when you put this into clinical practice.”
Shetty said that the AI system would help radiologists rather than replace them: “They each bring their own strength, it’s complementary. There are a number of cases where the radiologists catch something that the model misses, and vice versa. Bringing the two together could strengthen the overall results.”
