“`html
The emergence of more accessible artificial intelligence (AI) frameworks has revolutionized the domain of health assessments and medicine, with AI being utilized for diagnostic precision, customized treatment strategies, interpreting medical imagery, optimizing processes, facilitating remote patient surveillance, and much more.
Investigators from the eHealth Lab at Florida State University’s School of Information have been analyzing the role of AI as a resource to assist health care professionals in achieving more precise patient diagnostics. This progress holds the promise of enhancing treatment methodologies and improving patient results.
Senior Author and Director for FSU’s Institute for Successful Longevity Zhe He and Visiting Assistant Professor Balu Bhasuran are among the co-authors on the cross-institutional investigation. The research has already attracted considerable attention, with the publication being accessed over 3,000 times since its release in mid-March.
The manuscript, published in npj Digital Medicine, builds on FSU’s LabGenie project, a patient-engagement initiative designed to enhance older adults’ comprehension of lab test results.
The research group has been investigating the practicality of using large language models (LLMs), a kind of AI that learns from extensive amounts of text to respond to queries accurately, to support clinicians and bolster differential diagnosis precision and efficiency. Differential diagnosis (DDx) is a vital component in clinical decision-making, enabling health care providers to differentiate between conditions presenting similar symptoms.
“The AI-generated differential diagnosis is quite thorough in addressing all possible cases for patients,” He stated. “What this study illustrates is how AI can be leveraged as a tool to assist practitioners in making more enlightened decisions for their patients.”
The investigation involved employing the LLMs to create lists of the top one, five, and ten DDx for clinicians’ assessment. Researchers evaluated the precision and predictive capacity of the LLMs and studied how the inclusion of lab test results influenced their diagnostic accuracy.
“What this study illustrates is how AI can be harnessed as a resource to assist practitioners in making more enlightened decisions for their patients.”
– Zhe He, senior author and director for FSU’s Institute for Successful Longevity
The investigation assessed five LLMs — GPT-4, GPT-3.5, Llama-2-70b, Claude-2, and Mixtral-8x7B — utilizing clinical vignettes, or narrative patient-related scenarios, derived from 50 case studies. The results indicate that lab test information significantly enhances diagnostic accuracy, with GPT-4 attaining the top performance.
Specifically, GPT-4 reached 55% top one precision and 60% top 10 precision with lab information, with lenient accuracy reaching 80%. Lab exams, including liver function, metabolic/toxicology panels, and serology/immune tests, were generally interpreted accurately by the LLMs.
“When we prompted the model for the top differential diagnosis, most of these models were capable of pinpointing the patient’s precise diagnosis,” Bhasuran remarked. “That’s quite fascinating because it suggests that even for rare diseases, the model is able to make accurate predictions.”
The study aims to address well-recognized concerns often felt in health care environments from both provider and patient viewpoints. Accurate diagnosis is essential for effective patient management, directly impacting treatment choices and overall patient outcomes. Minimizing diagnostic errors helps streamline patient care, eliminating the necessity for redundant or excessive testing and ultimately reducing health care expenses through shorter hospital stays and fewer unnecessary procedures.
The research was backed by an Agency for Healthcare Research and Quality grant and partially supported by the University of Florida-Florida State University Clinical and Translational Science Award and the National Library of Medicine. Collaboration included Tampa General Hospital and coauthors from Florida State University, the National Library of Medicine, Emory University, University of South Florida, and University of North Texas Health Science Center. FSU Undergraduate Research Opportunity Program (UROP) students Angelique Deville, Hailey Thompson, Maggie Awad, and Yash Alva contributed by extracting key information for the case studies.
For further details, visit ehealthlab.cci.fsu.edu.
The article FSU researchers’ new study explores AI’s capacity to enhance differential diagnosis accuracy first appeared on Florida State University News.
“`