On Mammograms, Google's AI Plays Big Role in Breast Cancer AID

Google has created algorithms that will detect lung cancers on CT scans, an Eye disease in people with diabetes. Computers are trained to recognize patterns and interpret images, which will help in finding breast cancer on microscopic slides.

Samples are tested, where the diagnosis was already known, the system performed better than radiologists. On scans from the United States, the system produced a 9.4% axing in false negatives, where mammogram accidentally read as usual, and cancer was missed, which provide an underneath of 5.7% in amiss, where the scan is falsely judged unusual, but there is no sign of cancer. On mammograms that performed in Britain, the system also defeated the radiologists, reducing false negatives by 2.7% and flawed by 1.2%.

As per the National Breast Cancer Foundation, Breast cancer is the most common type of cancer, which is found in women and with a rate of one in eight women in the US developing the condition in her lifetime. Early detection and treatment are linked to better survival rates. AI is intended to help radiologists not to replace them. The combination of AI & Radiologists achieves the most accurate diagnostic results, which would help free up radiologist's time to focus more on patient care.

Google is worked with researchers from Northwestern University in Chicago and two British medical centers, Cancer Research Imperial Centre & Royal Surrey County Hospital. In the US, 33 million screening mammograms are performed each year and test misses 20% of breast cancers.

To train computer AI using mammograms from about 76,000 women in Britain and 15,000 in the US, whose diagnoses were already known. Then, computers are testing images on 25,000 women in Britain and 3,000 in the US and compared the system's performance with radiologists who had read the X-rays.

Researchers marked AI against six radiologists in the US, presenting 500 mammograms to be interpreted. Overall, AI again outperformed humans. But in some instances, AI missed cancer that all six radiologists found.

According to Lehman, who is also developing AI for mammograms, "next step in the research is radiologists try using the tool to practice in reading mammograms. In 1998, computer-aided detection, or CAD, came into use to help radiologists read mammograms. But it was found that CAD did not improve the doctor's precision and even made it worse. We can learn from the mistakes and do it better, and now adding AI, results become far more powerful and accurate."