Google announced that it has developed a new artificial intelligence algorithm for doctors to diagnose breast cancer. The algortima, which scans mammogram films, reduces false negatives by 9.4 percent. Today, due to false negatives, the cancer does not appear in the films taken by patients with breast cancer.
20 percent of patients with breast cancer are therefore unable to benefit from early diagnosis, whereas breast cancer can only be treated with early diagnosis.
Shravya Shetty, who conducted research on the subject at Google, said mammograms are very effective during the diagnosis, but false negatives and false positives pose a serious problem. Shetty added that the study followed the principles followed by radiologists.
The artificial intelligence, which examines the anonymizations of 3,000 women from 25,000 US from the United Kingdom, was first trained to scan X-ray images. She then detects changes in the breasts of 28,000 women. In the final stage, the algorithm’s estimate is controlled by the patient’s medical data.
Algortima reduced false negatives by 9.4 percent and false positives by 5.7 percent in the United States. In the United Kingdom, the algorithm reduced false negatives by 2.7 percent, while false positives decreased by 12 percent.
Christopher Kelly, who worked in Google research, said the algorithm performs better than radiologists in both the United States and the United Kingdom.
On the other hand, in the past, the algorithm has made mistakes. The cancer, which was overlooked by the algorithm, was detected by 6 US radiologists.
Google stressed that its project was not designed to tip radiologists out of work, but to help them:
Let’s say that with this new research, Google is strengthening its steps in the field of health. Google, which announced its partnership with Ascension in recent months following its acquisition of Fitbit, is determined to take a share of the healthcare sector. The company is preparing to produce a search engine that analyses millions of health records, and could allow new diagnostic methodologies to emerge in the coming years.