Japan uses artificial intelligence technology to identify early gastric cancer
- Jul 30, 2018 -

Japanese researchers have developed a method for identifying gastric cancer with artificial intelligence technology, which can detect early gastric cancer with high accuracy.

The Japanese Institute of Physical Chemistry released a communique recently, it is not easy to distinguish early gastric cancer and gastritis, and specialists may not be able to make accurate judgments through endoscopic images. To this end, researchers at the agency and researchers at the National Cancer Research Center of Japan decided to use artificial intelligence deep learning techniques to identify images of early gastric cancer.

Computer deep learning typically requires hundreds of thousands to millions of images for learning data. Because it is difficult to collect high-quality images of early gastric cancer, the researchers selected about 100 images of early gastric cancer and normal images, and randomly intercepted and expanded the data to generate about 360,000 images.

After the computer deeply studied a large amount of image data, the researchers used about 10,000 images that were not used for learning to test whether the computer can accurately diagnose early gastric cancer. As a result, it was found that the accuracy rate was 93.4% in the case of gastric cancer, and the accuracy was 83.6% in the case judged to be normal. In addition, after deep learning, the computer can determine whether it is cancerous or not, and can also point out the cancerous site.

The researchers said that the results of this study will help the discovery and treatment of early gastric cancer, and they will next study how to further improve the accuracy of computer recognition.