UPDATE : Monday, June 1, 2020
HOME Hospital
Seoul St. Mary’s Hospital develops AI diagnosing pneumoconiosis
  • By Lee Han-soo
  • Published 2018.01.15 17:50
  • Updated 2018.01.15 17:50
  • comments 0
Professor Myung Jun-pyo plans to additionally apply the research team’s AI system to forecast the growing diverse environmental diseases.

A research team at Seoul St. Mary’s Hospital has developed artificial intelligence (AI) system that can diagnose pneumoconiosis, the hospital said Monday.

Pneumoconiosis is an occupational lung disease caused by the inhalation of dust, which causes lung fibrosis. The accurate diagnosis of the disease is essential as the government pays industrial disaster compensation through at the Korea Workers’ Compensation and Welfare Service (KWCWS) after verifying the patient's dust exposure and image validation.

In Korea, about 10,000 people apply for screening each year. However, patients diagnosed with pneumoconiosis account for only 20 percent.

The low validation ratio has become an issue as hospitals had to spend extra personnel to examine patients, who did not have pneumoconiosis. Also, the change of the image confirmation procedures every three years caused disputes such as requests for reexamination and reconsideration as well as administrative litigation.

Such problems have caused the medical community to call for an accurate and consistent examination standard.

The St. Mary’s Hospital team led by Professor Myung Jun-pyo of the department of occupational medicine used pneumoconiosis confirmed images as the research group while using images of general examiners 60 or older who performed screening at the hospital as the contrast group. They also used KoSDI (Korea Pneumatic Standard Digital Image), a standard for validating pneumoconiosis produced and distributed by KWCWS, to improve the quality consistency of the images.

As a result, the AI system diagnosed pneumoconiosis with 95 percent accuracy. The system, a deep learning framework based on a CNN model, which is a technology that allows machines to learn vast amounts of data, studied more than 1,200 images of pneumoconiosis patients collected from May 2011 to March 2017.

“The team plans to conduct additional research to expand the role of the AI system in the screening of environmental diseases such as lung damage caused by asbestos and humidifier disinfectant with data accumulated at Seoul St. Mary's Hospital,” Professor Myung said.


<© Korea Biomedical Review, All rights reserved.>

Other articles by Lee Han-soo
iconMost viewed
Comments 0
Please leave the first comment.
Back to Top