GPT-4, Google Gemini Fall Short in Breast Imaging Classification

Use of publicly available large language models (LLMs) resulted in changes in breast imaging reports classification that could have a negative effect on patient management, according to a new international study published today in the journal Radiology, a journal of the Radiological Society of North America (RSNA). The study findings underscore the need to regulate these LLMs in scenarios that require high-level medical reasoning, researchers said.

LLMs are a type of artificial intelligence (AI) widely used today for a variety of purposes. In radiology, LLMs have already been tested in a wide variety of clinical tasks, from processing radiology request forms to providing imaging recommendations and diagnosis support.

Publicly available generic LLMs like ChatGPT (GPT 3.5 and GPT-4) and Google Gemini (formerly Bard) have shown promising results in some tasks. Importantly, however, they are less successful at more complex tasks requiring a higher level of reasoning and deeper clinical knowledge, such as providing imaging recommendations. Users seeking medical advice may not always understand the limitations of these untrained programs.

"Evaluating the abilities of generic LLMs remains important as these tools are the most readily available and may unjustifiably be used by both patients and non-radiologist physicians seeking a second opinion," said study co-lead author Andrea Cozzi, M.D., Ph.D., radiology resident and post-doctoral research fellow at the Imaging Institute of Southern Switzerland, Ente Ospedaliero Cantonale, in Lugano, Switzerland.

Dr. Cozzi and colleagues set out to test the generic LLMs on a task that pertains to daily clinical routine but where the depth of medical reasoning is high and where the use of languages other than English would further stress LLMs capabilities. They focused on the agreement between human readers and LLMs for the assignment of Breast Imaging Reporting and Data System (BI-RADS) categories, a widely used system to describe and classify breast lesions.

The Swiss researchers partnered with an American team from Memorial Sloan Kettering Cancer Center in New York City and a Dutch team at the Netherlands Cancer Institute in Amsterdam.

The study included BI-RADS classifications of 2,400 breast imaging reports written in English, Italian and Dutch. Three LLMs - GPT-3.5, GPT-4 and Google Bard (now renamed Google Gemini) - assigned BI-RADS categories using only the findings described by the original radiologists. The researchers then compared the performance of the LLMs with that of board-certified breast radiologists.

The agreement for BI-RADS category assignments between human readers was almost perfect. However, the agreement between humans and the LLMs was only moderate. Most importantly, the researchers also observed a high percentage of discordant category assignments that would result in negative changes in patient management. This raises several concerns about the potential consequences of placing too much reliance on these widely available LLMs.

According to Dr. Cozzi, the results highlight the need for regulation of LLMs when there is a highly likely possibility that users may ask them health-care-related questions of varying depth and complexity.

"The results of this study add to the growing body of evidence that reminds us of the need to carefully understand and highlight the pros and cons of LLM use in health care," he said. "These programs can be a wonderful tool for many tasks but should be used wisely. Patients need to be aware of the intrinsic shortcomings of these tools, and that they may receive incomplete or even utterly wrong replies to complex questions."

Cozzi A, Pinker K, Hidber A, Zhang T, Bonomo L, Lo Gullo R, Christianson B, Curti M, Rizzo S, Del Grande F, Mann RM, Schiaffino S.
BI-RADS Category Assignments by GPT-3.5, GPT-4, and Google Bard: A Multilanguage Study.
Radiology. 2024 Apr;311(1):e232133. doi: 10.1148/radiol.232133

Most Popular Now

Herefordshire and Worcestershire Health …

Herefordshire and Worcestershire Health and Care NHS Trust has successfully implemented Alcidion's Miya Precision platform to streamline bed management workflow across seven community hospitals in Worcestershire. The trust delivers community...

With Huge Patient Dataset, AI Accurately…

Scientists have designed a new artificial intelligence (AI) model that emulates randomized clinical trials at determining the treatment options most effective at preventing stroke in people with heart disease. The model...

A Shortcut for Drug Discovery

For most human proteins, there are no small molecules known to bind them chemically (so called "ligands"). Ligands frequently represent important starting points for drug development but this knowledge gap...

New Horizon Europe Funding Boosts Europe…

The European Commission has announced the launch of new Horizon Europe calls, with a substantial funding pool of over €112 million. These calls are aimed primarily at pioneering projects in...

Cleveland Clinic Study Finds AI can Deve…

Cleveland Clinic researchers developed an artficial intelligence (AI) model that can determine the best combination and timeline to use when prescribing drugs to treat a bacterial infection, based solely on...

New AI-Technology Estimates Brain Age Us…

As people age, their brains do, too. But if a brain ages prematurely, there is potential for age-related diseases such as mild-cognitive impairment, dementia, or Parkinson's disease. If "brain age...

Radboud University Medical Center and Ph…

Royal Philips (NYSE: PHG, AEX: PHIA), a global leader in health technology, and Radboud University Medical Center have signed a hospital-wide, long-term strategic partnership that delivers the latest patient monitoring...

GPT-4, Google Gemini Fall Short in Breas…

Use of publicly available large language models (LLMs) resulted in changes in breast imaging reports classification that could have a negative effect on patient management, according to a new international...

ChatGPT fails at heart risk assessment

Despite ChatGPT's reported ability to pass medical exams, new research indicates it would be unwise to rely on it for some health assessments, such as whether a patient with chest...

Study Shows ChatGPT Failed when Challeng…

With artificial intelligence (AI) poised to become a fundamental part of clinical research and decision making, many still question the accuracy of ChatGPT, a sophisticated AI language model, to support...

Virtual Reality Shows Promise in Fightin…

A new study published in JMIR Mental Health sheds light on the promising role of virtual reality (VR) in treating major depressive disorder (MDD). Titled "Examining the Efficacy of Extended...

AXREM and Highland Marketing Partner to …

AXREM represents member companies that collectively provide UK hospitals with most of their diagnostic medical imaging technology, and radiotherapy equipment. The association has seen substantial growth in recent years, with membership...