CSIRO researchers have developed a new method that helps artificial intelligence (AI) write more accurate chest X-ray reports. The technology uses the same information doctors rely on in real life.
The team trained the AI using over 46,000 patient cases from a major US hospital. The AI learned to combine X-ray images with emergency department data such as vital signs, medications, and clinical notes. Tests showed a 17 per cent improvement in diagnostic accuracy and closer alignment with expert radiologists.
Hospitals worldwide face shortages of radiologists. Faster and more reliable AI reporting could help reduce delays and ease pressure on clinical teams.
Dr Aaron Nicolson, lead author, said the AI works like a diagnostic detective. “When you combine what’s in the X-ray with what’s happening at the bedside, the AI gets more accurate, and much more useful,” he said.
The AI is being trialled at the Princess Alexandra Hospital in Brisbane. Researchers are comparing its reports with human radiologists to measure accuracy and usefulness. They are seeking additional hospitals to test the technology.
The code and dataset are freely available to researchers globally. This openness allows other teams to explore and improve AI-assisted diagnostics for chest X-rays.