Riedel Maximilian, Meyer Bastian, Kfuri Rubens Raphael, Riedel Caroline, Amann Niklas, Kiechle Marion, Riedel Fabian
Department of Gynecology and Obstetrics, TUM University Hospital, Technical University Munich (TU), Munich, Germany.
Institute of Computational Biology, Helmholtz Zentrum München, German Research Center for Environmental Health, Neuherberg, Germany.
Acta Obstet Gynecol Scand. 2025 Jul;104(7):1373-1381. doi: 10.1111/aogs.15123. Epub 2025 May 14.
The emergence of large language models heralds a new chapter in natural language processing, with immense potential for improving medical care and especially medical oncology. One recent and publicly available example is Generative Pretraining Transformer 4 (GPT-4). Our objective was to evaluate its ability to rephrase original surgical reports into simplified versions that are more comprehensible to patients. Specifically, we aimed to investigate and discuss the potential, limitations, and associated risks of using these simplified reports for patient education and information in gynecologic oncology.
We tasked GPT-4 with generating simplified versions from n = 20 original gynecologic surgical reports. Patients were provided with both their original report and the corresponding simplified version generated by GPT-4. Alongside these reports, patients received questionnaires designed to facilitate a comparative assessment between the original and simplified surgical reports. Furthermore, clinical experts evaluated the artificial intelligence (AI)-generated reports with regard to their accuracy and clinical quality.
The simplified surgical reports generated by GPT-4 significantly improved our patients' understanding, particularly with regard to the surgical procedure, its outcome, and potential risks. However, despite the reports being more accessible and relevant, clinical experts highlighted concerns about their lack of medical precision.
Advanced language models like GPT-4 can transform unedited surgical reports to improve clarity about the procedure and its outcomes. It offers considerable promise for enhancing patient education. However, concerns about medical precision underscore the need for rigorous oversight to safely integrate AI into patient education. Over the medium term, AI-generated, simplified versions of these reports-and other medical records-could be effortlessly integrated into standard automated postoperative care and digital discharge systems.
大语言模型的出现为自然语言处理开启了新篇章，在改善医疗护理尤其是医学肿瘤学方面具有巨大潜力。最近一个公开可用的例子是生成式预训练变换器4（GPT-4）。我们的目标是评估其将原始手术报告改写为患者更易理解的简化版本的能力。具体而言，我们旨在调查和讨论使用这些简化报告进行妇科肿瘤学患者教育和信息提供的潜力、局限性及相关风险。
我们让GPT-4从n = 20份原始妇科手术报告生成简化版本。向患者提供其原始报告以及GPT-4生成的相应简化版本。除这些报告外，患者还收到旨在便于对原始和简化手术报告进行比较评估的问卷。此外，临床专家评估了人工智能生成的报告在准确性和临床质量方面的情况。
GPT-4生成的简化手术报告显著提高了我们患者的理解程度，尤其是在手术过程、结果及潜在风险方面。然而，尽管这些报告更易于获取且相关性更强，但临床专家强调了对其缺乏医学精确性的担忧。
像GPT-4这样的先进语言模型可以将未经编辑的手术报告进行转化，以提高手术过程及其结果的清晰度。它在加强患者教育方面具有相当大的前景。然而，对医学精确性的担忧凸显了进行严格监督以将人工智能安全整合到患者教育中的必要性。从中期来看，人工智能生成的这些报告及其他医疗记录的简化版本可以轻松整合到标准的术后自动护理和数字出院系统中。
