More than 2.2 million U.S. workers are exposed to hazardous airborne dust from rock, sand or coal on the job, putting them at risk of permanent lung damage. Workers in industries such as mining, construction, engineered stone fabrication and foundries can develop serious conditions including black lung and silicosis after years of inhaling fine dust particles.

Regular chest X-rays can help detect lung damage before severe breathing problems develop. However, these scans must be reviewed by specially trained physicians known as B readers, and there are currently only about 200 certified B readers in the U.S. This shortage can contribute to screening backlogs and delays in identifying workers who may need protection from further exposure.

Researchers at Michigan State University have developed an artificial intelligence system designed to help physicians interpret these X-rays more quickly and consistently. Supported by a $600,000 grant from the National Institute for Occupational Safety and Health (NIOSH), the system was trained using a specialised dataset of scans from U.S. workers. The study was published in Occupational and Environmental Medicine.

“For workers in dusty environments, time is everything,” said Kenneth Rosenman, chief of MSU’s Division of Occupational and Environmental Medicine and a certified B reader. If lung disease goes undetected, he explained, workers may remain exposed to hazardous dust while additional scarring develops. The AI system is intended to provide physicians with an objective second opinion that could help identify disease earlier.

Early detection is particularly important because dust-related lung damage often develops gradually over 10 to 20 years. Workers may not realise they are affected until they experience serious breathing difficulties, and once lung tissue has been scarred, the damage cannot be reversed. Earlier identification can allow workers to move into lower-exposure roles and receive appropriate medical care.

Screening also helps employers determine whether workplace dust controls are effective and whether stronger ventilation or other protections are needed. Early diagnosis may also help affected workers obtain medical treatment and, where eligible, financial compensation. Coal miners, for example, require routine lung scans every five years, adding to the large volume of images specialists must review.

The MSU system demonstrated strong performance across four screening tasks. It was able to identify and clear about half of normal X-rays, potentially allowing specialists to devote more time to scans showing possible disease. The AI also achieved 91% accuracy in detecting the earliest signs of lung scarring, compared with an average of 77% among human readers. A colour map highlights areas where potential damage is detected for clinicians to review.

“Because early signs of lung damage are so subtle, even experienced human doctors often disagree on what they see,” said Ling Wang, associate professor in MSU’s Division of Occupational and Environmental Medicine. The researchers envision the technology as a second opinion rather than a replacement for specialists. The MSU team is now working with NIOSH to package the technology into an application that could eventually support wider use.

More information: Meiqi Liu et al, Pneumoconiosis screening and classification using deep learning models, Occupational and Environmental Medicine. DOI: 10.1136/oemed-2026-111040

Journal information: Occupational and Environmental Medicine Provided by Michigan State University