In 2019, the American Cancer Society stated an expected 268,600 new cases of invasive breast cancer and 41,760 deaths among women in the United States. However, amongst the 33 million screening mammograms performed each year, the tests are said to miss about 20% of the cancers, false positives being common. To be told you have cancer when you do not, or to have caught undetected instances of cancer are situations that are both harmful to patients. Unfortunately, mistakes that occur when reading mammograms are sometimes well outside the acceptable margins of normal human error, but a potential solution to this issue is Google’s latest artificial intelligence (AI) system.
The system was designed and trained to recognize cancer using mammograms from 15,000 women whose diagnoses were already known. The same system was then tested with images from 3,000 other women before comparing results to that of the radiologists’ who originally read the X-rays. Researchers were able to conclude that the AI system was more accurate. Additional test results also proved the system’s ability to outperform six radiologists when presented with 500 mammograms to be interpreted, as well as the ability to perform at the same standard as two doctors working together. In some instances, however, the AI system missed a cancer detection that all six radiologists found.
Globally, there are about 2 million new cases every year, with more than 500,000 deaths. This successful research study introduces the use of AI to enable accurate and efficient screening, as well as to reduce the time taken for diagnosis down to seconds. The need for only one radiologist will ease the pressure on the others’ workload and allow them to prioritize other important matters that would require a human touch.