Artificial Intelligence Program Predicts Cancer With 86 Percent Accuracy

An artificial intelligence program that scans thousands of human body cells was able to predict – with 86 percent accuracy – which would become cancerous.

The Yokohama, Japan AI study was able to detect colorectal cancer before benign tumors were able to become malignant. The program took microscopic images of a colorectal polyp, magnified it by 500 times and then cross-referenced the variations with more than 30,000 images. The database of images contained both pre-cancerous and cancerous cells, and was the first cross-referencing image research of its kind, Inverse first reported.

The AI-assisted system pulled off an impressive 86 percent prediction accuracy rate that was derived after assessing patients with a colorectal polyp diagnosis. The cancer detection process by the AI program occurs in under one second.

“Overall, 306 polyps were assessed real-time by using the AI-assisted system, providing a sensitivity of 94 percent specificity of 79 percent, accuracy of 86 percent, and positive and negative predictive values of 79 percent and 93 percent respectively, in identifying neoplastic changes,” wrote study authors Dr. Yuichi Mori of Showa University. The AI detection program is similar to the collection of face scans or other massive databases for image comparison. The before and after effect of this current AI research has great potential to identify specific cancer risks before they become malignant.

Dr. Mori presented he and his colleagues’ AI cancer prediction findings at the United European Gastroenterology conference going on this week in Barcelona, Spain.

Colorectal cancer is the second deadliest form behind lung cancer. Previous research has shown that the thin walls of the colon, rectum and intestine allow the cancerous cells to easily work themselves into the bloodstream and spread throughout the body.

Dr. Mori’s AI study aims to potentially increase colorectal cancer survival rates by creating a database of pre and post-cancerous cells that can even more accurately predict the most at-risk cells.

“We believe these results are acceptable for clinical application and our immediate goal is to obtain regulatory approval for the diagnostic system,” Mori said in a statement.