In brief: Over the past year or so, we've seen quite a few impressive advancements in the realm of health-related AI technology. Some AIs have proven capable of detecting heart conditions early, or even diagnosing ocular diseases with the accuracy of world-leading doctors. Now, Google is using AI to tackle breast cancer.
More specifically, the tech giant is using AI to improve breast cancer detection rates. In a blog post, Google notes that over 55,000 people in the United Kingdom alone are diagnosed with the deadly disease annually, and one in eight American women will likely develop it at some point in their lifetime.
These numbers probably won't come as a shock to most of our readers. We've all seen the countless breast cancer awareness campaigns run by non-profits and corporations alike over the years. However, despite how well-known the illness is, it's still remarkably difficult to treat. Your best chance of tackling any sort of cancer usually lies in early detection. In some cases, once the symptoms set in, cancer might already be too advanced to effectively treat.
Unfortunately, even early detection isn't foolproof. Indeed, current breast cancer screenings (called "mammograms") reportedly miss about 20 percent of cases. An 80 percent success rate is certainly better than nothing, but it's far from ideal.
Google's AI, though, has already proven capable of improving detection rates. To accomplish this feat, Google fed its AI 76,000 "de-identified" mammograms from UK women, and 15,000 from US women. Once the model was trained, it was put to the test in a separate data set, which contained mammograms from over 25,000 UK residents and 3,000 US citizens.
During this test, the AI managed to reduce false negatives by 9.4 percent (compared to average detection rates), and false positives by 5.7 percent. Google feels these results show promise, and -- if the tech is used in actual clinical settings -- could help doctors "reduce wait times and stress" for potential cancer patients down the line.
Whether or not Google's AI will indeed be put to the test in a real-world setting remains to be seen. There are plenty of hurdles for the AI to clear (both legal and scientific) before that day will come, but we'll keep you updated on the tech moving forward.
Masthead credit: Shutterstock