In Japan, AI was taught to detect liver diseases from chest X-rays

Scientists from Osaka University have developed artificial intelligence that can detect fatty liver disease - a condition where fat accumulates in the liver and can eventually lead to cirrhosis or cancer. The work is published in the journal Radiology Cardiothoracic Imaging (RCI).

Currently, ultrasound, CT, or MRI are typically used to diagnose this disease, but these are expensive examinations requiring specialized equipment. At the same time, chest X-rays are much more accessible and inexpensive examinations that many patients undergo. Part of the liver is also visible on X-rays, and as scientists have discovered, this is enough for AI to notice signs of the disease.

Researchers trained the model on 6,599 X-ray images from 4,414 patients and demonstrated high accuracy in detecting fatty hepatosis - the AUC indicator, which reflects the quality of diagnosis, was 0.82-0.83.

"Using regular X-rays to diagnose this disease can significantly simplify and reduce the cost of early detection of fatty liver disease," noted Professor Sawako Uchida-Kobayashi.

The scientists hope that the development will be used in real medical practice in the future.