Researchers at Severance Hospital have developed an artificial intelligence (AI) model that analyzes gastroscopic images to detect early gastric cancer and predict the depth of tumor invasion.

Professor Kim Ji-hyun

Professors Kim Ji-hyun and Yoon Hong-jin from the hospital and Selvas AI, a local AI company, participated in the development of the new software.

The team developed a new model based on the VGG-16, an AI model widely used for image classification, and trained it further by providing the AI with 800 early gastric cancer lesions through 11,539 endoscopic images. It confirmed a detection accuracy of 98.5 percent for early gastric cancer and an 85.1 percent accuracy regarding the depth of the tumor invasion.

“Early gastric cancer can be cured with endoscopic resection and does not require surgery, depending on the depth of tumor invasion,” Professor Kim said. “The predictive accuracy of the newly developed AI model will help us determine the diagnosis and treatment of early gastric cancer.”

Journal of Clinical Medicine published the result of the study.

“We are honored to have published research in an international journal by presenting a new methodology that applies AI technology to gastric cancer diagnosis and depth of tumor invasion," an official from Selvas AI said. “Through further research and development, we will become a leading company in the advancement of AI medical technology.”

Copyright © KBR Unauthorized reproduction, redistribution prohibited