Asan Medical Center’s AI model detects intestinal holes in newborns

The model achieved an external validation accuracy of 84.1%.

A research team from Asan Medical Center has developed an artificial intelligence (AI) interpretation model that analyses newborn X-ray images to detect intestinal perforation and identify the exact location of the lesion.

The model achieved an internal validation accuracy of 94.9% and an external validation accuracy of 84.1%.

The team designed a deep multitask learning model that can both determine the presence of perforation and highlight air-filled areas in the abdominal cavity on X-ray images.

In addition, data augmentation techniques were applied, allowing the model to learn from a wide range of imaging variations.

The team was led by Hee Mang Yoon of the Department of Radiology, Namkug Kim of the Department of Convergence Medicine, and Byong Sop Lee of the Division of Neonatology.