Australia’s national science agency, CSIRO, says it has developed what it describes as a world-first way to train artificial intelligence to write more accurate chest X‑ray reports by supplying the same bedside context clinicians use, rather than relying on images alone. The work comes as hospitals grapple with rising demand and persistent radiologist shortages.
Using more than 46,000 real‑world patient cases from a leading US hospital dataset, researchers trained a multimodal language model to produce detailed radiology reports. They report a 17 per cent improvement in diagnostic insights and closer alignment with expert radiologist reporting.
Instead of limiting algorithms to the scan and a brief referral, the team at CSIRO’s Australian e‑Health Research Centre combined imaging with emergency department information such as vital signs, medication histories and clinical notes.
“The AI is functioning as a diagnostic detective and we’re equipping it with more evidence,” said lead author Dr Aaron Nicolson. “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.”
Dr Nicolson presented the findings at the Association for Computational Linguistics conference in Vienna, Austria. “This is a practical, scalable way to help overworked clinical teams, reduce diagnostic delays, and ultimately improve outcomes for patients,” Dr Nicolson said.
Professor Ian Scott, Research Fellow at the University of Queensland Digital Health Centre and Clinical Consultant in AI at Metro South Hospital and Health Service, which has been involved in testing the technology, said there was strong potential in the approach. “For hard pressed radiologists confronting ever increasing workloads, we need this type of automated multimodal technology to reduce cognitive burden, improve workflows and allow timely and accurate reporting of chest X-rays for treating clinicians,” Professor Scott said.
The system is currently on trial at Brisbane’s Princess Alexandra Hospital to compare AI‑generated reports with those of human radiologists, and the team is seeking additional trial sites.
CSIRO says it has made its code and dataset freely available to researchers worldwide to encourage further innovation in AI‑assisted diagnostics.
