For pathologists, the hardest job is to determine if the patient has cancer. The diagnosis will determine their fate. But making an accurate diagnosis is very, very difficult.
It has been reported that the consensus of different doctors on the same pathological sample can be as low as 48%. So Google launched the Deep Learning AI project to help patients diagnose their condition.
The researchers said: "The patient will digitize multiple pathological samples, magnify 40 times, about 10 billion pixels of images. Imagine carefully checking thousands of sheets.
Tens of pixels of pictures, not every pixel. Without careful calculation, I understand how big the data is and complete it in a limited time."
Google has done this with advanced image recognition technology developed over the years and GoogLeNet. It was originally used to help autopilot cars identify roads and traffic signals. Now Google is using this technology for cancer diagnosis.
Google said: After targeted revisions, including training AI systems can view images at different magnifications (just like the pathologist's diagnosis process), we proved that you can train artificial intelligence to become a qualified pathologist, or its performance Better.
In an unrestricted pathological diagnosis, the accuracy rate of experienced physicians was 73%. After a short period of debugging, GoogLeNet achieved an accuracy of 89%.
bonito,australian bonito,smoked bonito,leaping bonito,black bonito
Zhoushan Boda Aquatic Products Co.,Ltd , https://www.baida-aquatic.com