Breast cancer is the number one form of cancer in women. Today, the mammogram is the most popular method of screening for and detecting the disease. Unfortunately, it requires radiologists to manually review the scans and look for cancerous areas. This leaves a large degree of human error in the process and leads to high percentages of missed diagnoses and false alarms.
Now, Google is pioneering an AI from its Health division that scans mammogram images and identifies breast cancer. It is able to pick up on positive readings even when experienced radiologists miss them. The team hopes that it will help improve the accuracy of breast cancer diagnoses and help eliminate missed cases.
Keen AI Vision
According to the American Cancer Society, one in eight women will develop breast cancer. Fortunately, modern treatments for the disease help more women beat their cancer and live longer than ever before. Still, the key to effective treatment is early detection.
That’s why recommendations suggest that women get a mammogram annually or every two years. Sadly, false diagnoses increase the anxiety surrounding the procedure. More than 50 percent of women who have screenings over the course of 10 years will have a false alarm. This typically leads to a painful biopsy to confirm or negate the findings.
Scientists and doctors obviously want to avoid this problem. Google Health’s AI is currently being tested as a way to help doctors identify breast cancer in mammogram screenings. The project is a collaboration with Northwestern University and the Cancer Research UK Imperial Centre. The research and results of the study are published in the journal Nature.
Google trained its AI with the anonymous scans of more than 28,000 women from the U.K. and U.S. It is able to spot instances of breast cancer better than radiologists in some cases. However, it also failed to flag several positive scans that were diagnosed unanimously by a panel of the same professionals.
Although that seems like a disappointment researchers actually see it as a positive. Currently, mammogram scans need to be double-verified by two radiologists. The team hopes that one day the AI can replace one set of human eyes. They envision it working in tandem with human doctors to identify cancer with more accuracy and more certainty. The AI could spot cancer with more precision while human radiologists act as a safeguard against missed cases.
Dr. Mozziyar Etemadi, a co-author of the study, says, “In some examples, the human outperforms the AI and in others, it’s the opposite. But the ultimate goal will be to find the best way to combine the two—the magic of the human brain isn’t going anywhere any time soon.”
The early results achieved with this AI are certainly impressive. However, more work is needed before it can be applied to a clinical setting.
Dominic King, the U.K. lead for Google Health, said via a statement that, “Further testing, clinical validation and regulatory approvals are required before this could start making a difference for patients, but we’re committed to working with our partners towards this goal.”
Should Google Health’s AI eventually make it into clinical practice it would be a huge step for the screening and early detection of breast cancer.
Learn more about the study via this video from Northwestern University in the player below: