AI sharpens diagnostic precision in hospitals amid patient privacy dilemmas

The use of artificial intelligence (AI) for the interpretation of chest x-rays in emergency departments has been shown to improve diagnostic accuracy by 5.9%, according to new research from the Australian Institute of Health Innovation (AIHI) at Macquarie University.

The study, led by Professor Farah Magrabi, AIHI’s Professor of AI and Patient Safety, involved 200 emergency department doctors who assessed 18 clinical scenarios with and without AI assistance. The findings reveal that integrating AI into radiological review significantly enhances both diagnostic precision and clinical decision-making.

“The use of AI to gain an on-the-spot interpretation of a chest x-ray will support clinicians to ensure highly critical diagnosis is not missed or treatment delayed in the emergency department,” said Professor Magrabi.

Beyond diagnostic accuracy, AI assistance also improved treatment and patient management decisions by 3.2%. The research is among only a few global studies to demonstrate measurable improvements in clinician performance resulting from AI-assisted decision-making.

Professor Michael Dinh, founder and director of the Royal Prince Alfred Hospital Green Light Institute for Emergency Care, co-authored the paper and participated in the project.

“More than half of all patients admitted to emergency departments are ordered a chest x-ray, but as formal radiology reporting can be delayed by hours, clinicians are sometimes under pressure to interpret results themselves in order to inform urgent clinical decisions. This study has shown that AI can support this decision-making,” Professor Dinh said.

He added, “Having an AI supported interpretation of the chest x-ray is a game changer for clinicians – enabling them to make on-the-spot decisions about diagnosis and support, ensuring patients receive faster and more appropriate care.”

These findings arrive a year after Harrison.ai launched its Annalise.ai diagnostic tool, which employs AI and large language models (LLMs) to analyse chest x-rays. The system was trained using de-identified patient data from the IMED Radiology Network, comprising around 250 clinics across Australia.

Following the launch, questions were raised about whether patient consent had been obtained for the use of their x-rays in AI training. An investigation by the Office of the Australian Privacy Commissioner found no breach, concluding that adequate deidentification likely ensured compliance under the Privacy Act, as sufficiently de-identified data does not meet the definition of “personal information”.

“There is, however, still a significant risk of reidentification based on the combination of characteristics taken from an individual patient’s data that may be used in training the AI. This risk is increased where multiple datasets containing data from the same patient are used together,” explained Andrew Lowe, Principal at Spruson & Ferguson.

The application of AI in emergency settings is particularly valuable given the mounting pressure on hospital departments. Professor Magrabi highlighted the potential for AI to strengthen diagnostic support in regional and remote hospitals with fewer senior consultants.

“While AI is being increasingly used in healthcare, this is one of a few studies to demonstrate the impact of AI on decision-making. Improving the accuracy of clinical interpretation is important, as is giving clinicians support to avoid misinterpretation of findings that can result in misdiagnosis and delayed treatment leading to poorer patient outcomes and more time in the emergency department,” she said.

Dr Mark Phillips, Chief Clinical Officer at Harrison.ai, said the results reinforce the clinical value of integrating AI into emergency workflows.

“We’re already working with emergency departments across Australia, Asia, and Europe, and this study reinforces the value of AI in supporting clinicians on the front line of care,” Dr Phillips said.

The study was funded by the Digital Health Cooperative Research Centre (DHCRC). CEO Annette Schmiede said, “Our partnership with Harrison.ai has supported the development of this world-class AI tool. Harrison.ai is an Australian success story, a homegrown company translating cutting-edge digital health research into clinical practice.

“By supporting clinicians with rapid and accurate interpretation of chest x-rays, this technology demonstrates how AI can transform emergency medicine, improve patient outcomes, and help deliver safer, more efficient care.”

In reimagining healthcare across the entire patient journey, Health Industry Hub^TM is the only one-stop-hub uniting the diversity of the Pharma, MedTech, Diagnostics & Biotech sectors to inspire meaningful change.

The Health Industry Hub^TM content is copyright protected. Access is available under individual user licenses. Please click here to subscribe and visit T&Cs here.