Somewhere in the not so distance future, computers could help doctors diagnose diseases much more quickly than they can today. In fact, researchers from Beth Israel Deaconess Medical Center (BIDMC) and Harvard Medical School (HMS) have been working on a way to train artificial intelligence (AI) to read and interpret pathology images that doctors use to look for signs of cancer.
Andrew Beck from BIDMC explains that the "method is based on deep learning." It is the method commonly used to train AI to recognize images, speech patterns, and objects. During a demonstration at the annual International Symposium of Biomedical Imaging, they were able to show how effective their training was. The AI was able to look for breast cancer in a series of images of lymph nodes.
The team started the training process by showing it hundreds of slides that were marked to indicate cancerous cells and normal cells. They then looked at the data and were able to identify which slides it was having trouble with and which ones were easy. From there, they fed it more and more difficult samples. AI was accurate 92% of the time, though it still has some room to grow because human pathologists are accurate 96% of the time.
Still, it is very promising.
Beck said that it is exciting and when they combined a pathologist with the creation, the results were 99.5% accurate.
Don't Miss: Incredible Pokemon Gifts
He added: "Our results in the ISBI competition show that what the computer is doing is genuinely intelligent and that the combination of human and computer interpretations will result in more precise and more clinically valuable diagnoses to guide treatment decisions."