Home Medizin Das KI-System kann Augenärzten bei der Diagnose und Behandlung von Glaukom und Netzhauterkrankungen ebenbürtig sein oder diese übertreffen

Das KI-System kann Augenärzten bei der Diagnose und Behandlung von Glaukom und Netzhauterkrankungen ebenbürtig sein oder diese übertreffen

von NFI Redaktion

According to a study by the New York Eye and Ear Infirmary of Mount Sinai, an artificial intelligence (AI) system, a large language model (LLM), may be equal to or even superior to human eye doctors in diagnosing and treating patients with glaucoma and retinal diseases (NYEE).

The provocative study, published on February 22 in JAMA Ophthalmology, suggests that advanced AI tools trained on vast amounts of data, text, and images could play a crucial role in supporting eye doctors in diagnosing and treating cases of glaucoma and retinal diseases, which affect millions of patients.

The study compared the knowledge of eye doctors with the abilities of the latest generation AI system, GPT-4 (Generative Pre-Training–Model 4) from OpenAI, which aims to mimic human performance. In medicine, advanced AI tools are seen as potentially revolutionary diagnostic and treatment tools due to the accuracy and completeness of their LLM-generated responses. Ophthalmology, with its high prevalence of complex patients, could be a particularly fertile field for AI and provide specialists with more time to practice evidence-based medicine.

The performance of GPT-4 in our study was quite insightful. We recognized the immense potential of this AI system from the moment we started testing, and we were fascinated to observe that GPT-4 could not only support but in some cases even surpass or outperform the expertise of experienced eye specialists.“

Andy Huang, MD, Assistant Professor of Ophthalmology at NYEE and lead author of the study

For the human side of the study, the Mount Sinai team recruited 12 treating specialists and three senior residents from the Department of Ophthalmology at the Icahn School of Medicine at Mount Sinai. A basic set of 20 questions (10 each for glaucoma and retina) from the list of frequently asked questions by patients from the American Academy of Ophthalmology was randomly selected, along with 20 unidentified patient cases from eye clinics associated with Mount Sinai. Subsequently, the responses of both the GPT-4/AI system and human specialists were statistically analyzed and evaluated for accuracy and thoroughness using a Likert scale commonly used in clinical research to assess responses.

The results showed that AI was able to match or even exceed human specialists in both the accuracy and completeness of their medical advice and assessments. Specifically, AI demonstrated superior performance in answering glaucoma questions and providing case management advice, while showing a more balanced outcome in retina questions, where AI matched humans in accuracy but exceeded them in completeness.

„AI was particularly surprising in its competence in handling glaucoma and retinal patient cases, matching the accuracy and completeness of diagnoses and treatment recommendations made by human doctors in a clinical note format,“ says Louis R. Pasquale, MD, FARVO, Vice Chair for Ophthalmology Research at the Department of Ophthalmology and senior author of the study. „Just as the AI application Grammarly can help us become better writers, GPT-4 can provide valuable insights on how we can become better clinicians, especially in documenting patient examination results.“

While Dr. Huang emphasizes that further testing is necessary, he believes that this work points to a promising future for AI in ophthalmology. „It could serve as a reliable assistant for eye doctors, providing diagnostic support and potentially easing their workload, especially in complex cases or high patient volume areas,“ he explains. „For patients, the integration of AI into general ophthalmology practice could lead to quicker access to expert advice and more informed decision-making to guide their treatment.“


Mount Sinai Health System

Journal Reference:

Huang, AS, et al. (2024). Evaluation of a large language model’s responses to glaucoma and retina management questions and cases. JAMA Ophthalmology. doi.org/10.1001/jamaophthalmol.2023.6917.

Related Posts

Adblock Detected

Please support us by disabling your AdBlocker extension from your browsers for our website.