Research News

Lifesaving deep learning model developed using standard radiographs

AI decision-making process with chest X-ray images

Radiographs of the heart and lungs also capture parts of the liver, allowing for deep learning models to detect fatty liver disease.

Credit: Osaka Metropolitan University

Fatty liver disease, caused by the accumulation of fat in the liver, is estimated to affect one in four people worldwide. If left untreated, it can lead to serious complications, such as cirrhosis and liver cancer, making it crucial to detect early and initiate treatment.

Currently, standard tests for diagnosing fatty liver disease include ultrasounds, CTs, and MRIs, which require costly specialized equipment and facilities. In contrast, chest X-rays are performed more frequently, are relatively inexpensive, and involve low radiation exposure. Although this test is primarily used to examine the condition of the lungs and heart, it also captures part of the liver, making it possible to detect signs of fatty liver disease. However, the relationship between chest X-rays and fatty liver disease has rarely been a subject of in-depth study.

Therefore, a research group led by Associate Professor Sawako Uchida-Kobayashi and Associate Professor Daiju Ueda at Osaka Metropolitan University’s Graduate School of Medicine developed an AI model that can detect the presence of fatty liver disease from chest X-ray images.

In this retrospective study, a total of 6,599 chest X-ray images containing data from 4,414 patients were used to develop an AI model utilizing controlled attenuation parameter (CAP) scores. The AI model was verified to be highly accurate, with the area under the receiver operating characteristic curve (AUC) ranging from 0.82 to 0.83.

“The development of diagnostic methods using easily obtainable and inexpensive chest X-rays has the potential to improve fatty liver detection. We hope it can be put into practical use in the future,” stated Professor Uchida-Kobayashi.

Paper Information

Journal: Radiology: Cardiothoracic Imaging
Title: Performance of a Chest Radiograph-based Deep Learning Model for Detecting Hepatis Steatosis
DOI: 10.1148/ryct.240402
Authors:  Daiju Ueda, Sawako Uchida-Kobayashi, Akira Yamamoto, Shannon L. Walston, Hiroyuki Motoyama, Hideki Fujii, Toshio Watanabe, Yukio Miki, MD, Norifumi Kawada
Published: 20 June 2025
URL: https://doi.org/10.1148/ryct.240402

Contact

Sawako Uchida-Kobayashi
Graduate School of Medicine
Email: sawako[at]omu.ac.jp

*Please change [at] to @.