[1]王璐,丁慕菲,周鹤,等.医学大语言模型的研发与应用系统综述[J].智能系统学报,2025,20(6):1295-1303.[doi:10.11992/tis.202410020]
WANG Lu,DING Mufei,ZHOU He,et al.Developing and employing large language models in medicine[J].CAAI Transactions on Intelligent Systems,2025,20(6):1295-1303.[doi:10.11992/tis.202410020]
点击复制
WANG Lu,DING Mufei,ZHOU He,et al.Developing and employing large language models in medicine[J].CAAI Transactions on Intelligent Systems,2025,20(6):1295-1303.[doi:10.11992/tis.202410020]
医学大语言模型的研发与应用系统综述
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
20
期数:
2025年第6期
页码:
1295-1303
栏目:
综述
出版日期:
2025-11-05
- Title:
- Developing and employing large language models in medicine
- Keywords:
- chatbot; artificial intelligence; large language models; ChatGPT; health care; clinical diagnosis; medical consultation; medical informatics
- 分类号:
- TP18; R319
- 摘要:
- 自2022年11月ChatGPT（chat generative pre-trained Transformer）问世以来，针对医学应用场景的大语言模型（large language models, LLMs）相关研究逐渐成为热点。然而当前缺乏对医学大语言模型研发以及应用现状的系统分析。为了更好地理解这些专门为医学领域设计的LLMs并评估其应用价值，本综述系统分析了截止至2024年6月11日，在PubMed、Google Scholar、arXiv、bioXiv和medRxiv等数据库中发表的为医学领域开发的专有LLMs，同时对LLMs在临床应用场景中的相关应用研究进行了梳理。研究结果表明，当前共计129项研究提出了基于医学相关语料研发的医学LLMs，而基于LLMs在临床应用场景中的应用涵盖了LLMs对医疗咨询的回应、不同模型间的比较、与专业医生的性能对比，以及医疗从业相关人员对LLMs的观点等4类研究内容。综述结果表明，通用型LLMs，如ChatGPT、GPT-4等在生成医疗记录时的准确性和完整性较高，而专门针对某些疾病所研发的LLMs则更擅长回答特定病症的问题，尽管它们的答复在全面性方面可能有所欠缺。医疗专家在辨别LLMs生成的文本与人类医生的文本时可能面临困难，但LLMs对重复提问的回复存在变异性。此外，从医学伦理角度看，LLMs在易读性和可能涉及种族及地域偏见的传播方面存在挑战，而且缺乏从患者或医疗保险提供商视角对LLMs可信度和责任等问题进行评估的研究。
- Abstract:
- Since the introduction of ChatGPT (chat generative pre-trained Transformer) in November 2022, studies related to large language models (LLMs) for medical applications are increasing; however, a systematic review of this field is lacking. This review covered studies indexed in PubMed, Google Scholar, arXiv, bioXiv, and medRxiv up until June 31, 2024, and identified 129 medical LLMs. LLMs were evaluated in clinical contexts, including their responses to medical queries, performance comparison, and specialist evaluation. The results revealed that general-purpose LLMs, such as ChatGPT and GPT-4, demonstrate better accuracy in generating medical records, whereas disease-specific LLMs excel in niche areas but may lack comprehensiveness. Challenges include variability in responses, readability issues, and biases, with few studies on LLM trustworthiness from patient or insurance perspectives.
- 参考文献/References:
-
[1] OpenAI. Introducing ChatGPT[EB/OL]. (2022-11-30)[2025-08-29]. https://openai.com/index/chatgpt.
[2] BISWAS S S. Role of chat GPT in public health[J]. Annals of biomedical engineering, 2023, 51(5): 868-869.
[3] SHEN Yongliang, SONG Kaitao, TAN Xu, et al. HuggingGPT: solving ai tasks with chatgpt and its friends in hugging face[J]. Advances in neural information processing systems, 2023, 36: 1-27
[4] SINGHAL K, TU T, GOTTWEIS J, et al. Towards expert-level medical question answering with large language models[EB/OL]. (2023-05-16)[2025-08-29]. https://arxiv.org/abs/2305.09617.
[5] YANG Xi, CHEN A, POURNEJATIAN N, et al. GatorTron: a large clinical language model to unlock patient information from unstructured electronic health records[EB/OL]. (2022-12-16)[2025-08-29]. https://arxiv.org/abs/2203.03540.
[6] HUANG Zhi, BIANCHI F, YUKSEKGONUL M, et al. A visual-language foundation model for pathology image analysis using medical Twitter[J]. Nature medicine, 2023, 29(9): 2307-2316.
[7] TIU E, TALIUS E, PATEL P, et al. Expert-level detection of pathologies from unannotated chest X-ray images via self-supervised learning[J]. Nature biomedical engineering, 2022, 6(12): 1399-1406.
[8] YALAMANCHILI A, SENGUPTA B, SONG J, et al. Quality of large language model responses to radiation oncology patient care questions[J]. JAMA network open, 2024, 7(4): e244630.
[9] CUI Yiming, YANG Ziqiang, YAO Xin. Efficient and effective text encoding for Chinese llama and alpaca[EB/OL]. (2024-02-23)[2025-08-29]. https://arxiv.org/abs/2304.08177.
[10] YANG Songhua, ZHAO Hanjie, ZHU Senbin, et al. Zhongjing: enhancing the Chinese medical capabilities of large language model through expert feedback and real-world multi-turn dialogue[J]. Proceedings of the AAAI conference on artificial intelligence, 2024, 38(17): 19368-19376.
[11] PIERI S, MULLAPPILLY S S, KHAN F S, et al. BiMediX: bilingual medical mixture of experts LLM[EB/OL]. (2024-12-10)[2025-08-29]. https://arxiv.org/abs/2402.13253.
[12] SUKEDA I, SUZUKI M, SAKAJI H, et al. JMedLoRA: medical domain adaptation on Japanese large language models using instruction-tuning[EB/OL]. (2023-12-01)[2025-08-29]. https://arxiv.org/abs/2310.10083.
[13] DEVLIN J, CHANG Mingwei, LEE K, et al. BERT: pre-training of deep bidirectional Transformers for language understanding[EB/OL]. (2019-05-24)[2025-08-29]. https://arxiv.org/abs/1810.04805.
[14] LIN Zeming, AKIN H, RAO R, et al. Evolutionary-scale prediction of atomic-level protein structure with a language model[J]. Science, 2023, 379(6637): 1123-1130.
[15] TOUVRON H, LAVRIL T, IZACARD G, et al. LLaMA: open and efficient foundation language models[EB/OL]. (2023-02-27)[2025-08-29]. https://arxiv.org/abs/2302.13971.
[16] BOLTON E, VENIGALLA A, YASUNAGA M, et al. BioMedLM: A 2.7B parameter language model trained on biomedical text[EB/OL]. (2024-03-27)[2025-08-29]. https://arxiv.org/abs/2403.18421.
[17] LEE J, YOON W, KIM S, et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining[J]. Bioinformatics, 2020, 36(4): 1234-1240.
[18] JIANG L Y, LIU X C, NEJATIAN N P, et al. Health system-scale language models are all-purpose prediction engines[J]. Nature, 2023, 619(7969): 357-362.
[19] NIJKAMP E, RUFFOLO J A, WEINSTEIN E N, et al. ProGen2: exploring the boundaries of protein language models[J]. Cell systems, 2023, 14(11): 968-978.
[20] Facebook Research. ESM (evolutionary scale modeling): ESMFold code and model release[EB/OL]. (2024-08-01)[2025-08-29]. https://github.com/facebookresearch/esm.
[21] LUO Renqian, SUN Liai, XIA Yingce, et al. BioGPT: generative pre-trained transformer for biomedical text generation and mining[J]. Briefings in bioinformatics, 2022, 23(6): bbac409.
[22] HAN Tianyu, ADAMS L C, PAPAIOANNOU J M, et al. MedAlpaca: an open-source collection of medical conversational AI models and training data[J]. (2023-04-14)[2025-08-29]. https://arxiv.org/abs/2304.08247.
[23] GE Jin, SUN S, OWENS J, et al. Development of a liver disease-specific large language model chat interface using retrieval-augmented generation[J]. Hepatology, 2024, 80(5): 1158-1168.
[24] XU Xuhai, YAO Bingsheng, DONG Yuanzhe, et al. Leveraging large language models for mental health prediction via online text data[EB/OL]. (2024-01-28)[2025-08-29]. https://arxiv.org/abs/2307.14385.
[25] ZHOU Yukun, CHIA M A, WAGNER S K, et al. A foundation model for generalizable disease detection from retinal images[J]. Nature, 2023, 622(7981): 156-163.
[26] PENG Cheng, YANG Xi, CHEN Aokun, et al. A study of generative large language model for medical research and healthcare[J]. NPJ digital medicine, 2023, 6(1): 210.
[27] LU M Y, CHEN B, WILLIAMSON D F K, et al. A visual-language foundation model for computational pathology[J]. Nature medicine, 2024, 30(3): 863-874.
[28] CHEN R J, DING Tong, LU M Y, et al. Towards a general-purpose foundation model for computational pathology[J]. Nature medicine, 2024, 30(3): 850-862.
[29] PAN A, MUSHEYEV D, BOCKELMAN D, et al. Assessment of artificial intelligence chatbot responses to top searched queries about cancer[J]. JAMA oncology, 2023, 9(10): 1437-1440.
[30] HAVER H L, AMBINDER E B, BAHL M, et al. Appropriateness of breast cancer prevention and screening recommendations provided by ChatGPT[J]. Radiology, 2023, 307(4): e230424.
[31] AYERS J W, ZHU Z, POLIAK A, et al. Evaluating artificial intelligence responses to public health questions[J]. JAMA network open, 2023, 6(6): e2317517.
[32] PUGLIESE N, WAI-SUN WONG V W, SCHATTENBERG J M, et al. Accuracy, reliability, and comprehensibility of ChatGPT-generated medical responses for patients with nonalcoholic fatty liver disease[J]. Clin gastroenterol hepatol, 2024, 22(4): 886-889.
[33] HENSON J B, GLISSEN BROWN J R, LEE J P, et al. Evaluation of the potential utility of an artificial intelligence chatbot in gastroesophageal reflux disease management[J]. The American journal of gastroenterology, 2023, 118(12): 2276-2279.
[34] FINK M A, BISCHOFF A, FINK C A, et al. Potential of ChatGPT and GPT-4 for data mining of free-text CT reports on lung cancer[J]. Radiology, 2023, 308(3): e231362.
[35] TAYEBI ARASTEH S, HAN Tianyu, LOTFINIA M, et al. Large language models streamline automated machine learning for clinical studies[J]. Nature communications, 2024, 15(1): 1603.
[36] YAN Chao, GRABOWSKA M E, DICKSON A L, et al. Leveraging generative AI to prioritize drug repurposing candidates for Alzheimer’s disease with real-world clinical validation[J]. NPJ digital medicine, 2024, 7(1): 46.
[37] UEDA D, MITSUYAMA Y, TAKITA H, et al. ChatGPT’s diagnostic performance from patient history and imaging findings on the diagnosis please quizzes[J]. Radiology, 2023, 308(1): e231040.
[38] KUNG T H, CHEATHAM M, MEDENILLA A, et al. Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models[J]. PLoS digital health, 2023, 2(2): e0000198.
[39] MIHALACHE A, HUANG R S, POPOVIC M M, et al. Performance of an upgraded artificial intelligence chatbot for ophthalmic knowledge assessment[J]. JAMA ophthalmology, 2023, 141(8): 798-800.
[40] BHAYANA R, KRISHNA S, BLEAKNEY R R. Performance of ChatGPT on a radiology board-style examination: insights into current strengths and limitations[J]. Radiology, 2023, 307(5): e230582.
[41] ALI R, TANG O Y, CONNOLLY I D, et al. Performance of ChatGPT and GPT-4 on neurosurgery written board examinations[J]. Neurosurgery, 2023, 93(6): 1353-1365.
[42] SCHUBERT M C, WICK W, VENKATARAMANI V. Performance of large language models on a neurology board-style examination[J]. JAMA network open, 2023, 6(12): e2346721.
[43] SUCHMAN K, GARG S, TRINDADE A J. Chat generative pretrained transformer fails the multiple-choice American college of gastroenterology self-assessment test[J]. The American journal of gastroenterology, 2023, 118(12): 2280-2282.
[44] GOODMAN R S, RANDALL PATRINELY J, STONE C A Jr, et al. Accuracy and reliability of chatbot responses to physician questions[J]. JAMA network open, 2023, 6(10): e2336483.
[45] ZACK T, LEHMAN E, SUZGUN M, et al. Assessing the potential of GPT-4 to perpetuate racial and gender biases in health care: a model evaluation study[J]. The lancet digital health, 2024, 6(1): e12-e22.
[46] MCGOWAN A, GUI Yunlai, DOBBS M, et al. ChatGPT and Bard exhibit spontaneous citation fabrication during psychiatry literature search[J]. Psychiatry research, 2023, 326: 115334.
[47] BENARY M, WANG X D, SCHMIDT M, et al. Leveraging large language models for decision support in personalized oncology[J]. JAMA network open, 2023, 6(11): e2343689.
[48] MUKHERJEE P, HOU B, LANFREDI R B, et al. Feasibility of using the privacy-preserving large language model vicuna for labeling radiology reports[J]. Radiology, 2023, 309(1): e231147.
[49] RAHSEPAR A A, TAVAKOLI N, KIM G H J, et al. How AI responds to common lung cancer questions: ChatGPT vs Google Bard[J]. Radiology, 2023, 307(5): e230922.
[50] SANDMANN S, RIEPENHAUSEN S, PLAGWITZ L, et al. Systematic analysis of ChatGPT, Google search and Llama 2 for clinical decision support tasks[J]. Nature communications, 2024, 15(1): 2050.
[51] WU Shaohong, TONG Wenjuan, LI Mingde, et al. Collaborative enhancement of consistency and accuracy in US diagnosis of thyroid nodules using large language models[J]. Radiology, 2024, 310(3): e232255.
[52] SAVAGE T, NAYAK A, GALLO R, et al. Diagnostic reasoning prompts reveal the potential for large language model interpretability in medicine[J]. NPJ digital medicine, 2024, 7(1): 20.
[53] AMIN K S, DAVIS M A, DOSHI R, et al. Accuracy of ChatGPT, Google Bard, and microsoft Bing for simplifying radiology reports[J]. Radiology, 2023, 309(2): e232561.
[54] LI D J, KAO Yuchen, TSAI S J, et al. Comparing the performance of ChatGPT GPT-4, Bard, and Llama-2 in the Taiwan Psychiatric Licensing Examination and in differential diagnosis with multi-center psychiatrists[J]. Psychiatry and clinical neurosciences, 2024, 78(6): 347-352.
[55] LIM Z W, PUSHPANATHAN K, YEW S M E, et al. Benchmarking large language models’ performances for myopia care: a comparative analysis of ChatGPT-3.5, ChatGPT-4.0, and Google Bard[J]. eBioMedicine, 2023, 95: 104770.
[56] OMIYE J A, LESTER J C, SPICHAK S, et al. Large language models propagate race-based medicine[J]. NPJ digital medicine, 2023, 6(1): 195.
[57] GERTZ R J, BUNCK A C, LENNARTZ S, et al. GPT-4 for automated determination of radiological study and protocol based on radiology request forms: a feasibility study[J]. Radiology, 2023, 307(5): e230877.
[58] GARCIA P, MA S P, SHAH S, et al. Artificial intelligence-generated draft replies to patient inbox messages[J]. JAMA network open, 2024, 7(3): e243201.
[59] BERNSTEIN I A, ZHANG Y V, GOVIL D, et al. Comparison of ophthalmologist and large language model chatbot responses to online patient eye care questions[J]. JAMA network open, 2023, 6(8): e2330320.
[60] DECKER H, TRANG K, RAMIREZ J, et al. Large language model-based chatbot vs surgeon-generated informed consent documentation for common procedures[J]. JAMA network open, 2023, 6(10): e2336997.
[61] RAU A, RAU S, ZOELLER D, et al. A context-based chatbot surpasses trained radiologists and generic ChatGPT in following the ACR appropriateness guidelines[J]. Radiology, 2023, 308(1): e230970.
[62] AYERS J W, POLIAK A, DREDZE M, et al. Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum[J]. JAMA internal medicine, 2023, 183(6): 589-596.
[63] LEE T C, STALLER K, BOTOMAN V, et al. ChatGPT answers common patient questions about colonoscopy[J]. Gastroenterology, 2023, 165(2): 509-511.
[64] BLEASE C, WORTHEN A, TOROUS J. Psychiatrists’ experiences and opinions of generative artificial intelligence in mental healthcare: an online mixed methods survey[J]. Psychiatry research, 2024, 333: 115724.
[65] EPPLER M, GANJAVI C, RAMACCIOTTI L S, et al. Awareness and use of ChatGPT and large language models: a prospective cross-sectional global survey in urology[J]. European urology, 2024, 85(2): 146-153.
[66] JIN J Q, DOBRY A S. ChatGPT for healthcare providers and patients: practical implications within dermatology[J]. Journal of the American Academy of Dermatology, 2023, 89(4): 870-871.
[67] YOUNG J N, O’HAGAN R, POPLAUSKY D, et al. The utility of ChatGPT in generating patient-facing and clinical responses for melanoma[J]. Journal of the American Academy of Dermatology, 2023, 89(3): 602-604.
[68] BENJAMENS S, DHUNNOO P, MESK? B. The state of artificial intelligence-based FDA-approved medical devices and algorithms: an online database[J]. NPJ digital medicine, 2020, 3: 118.
[69] BROWN S. Partnerships between health authorities and Amazon Alexa raise many possibilities: and just as many questions[J]. Canadian medical association journal, 2019, 191(41): E1141-E1142.
[70] SCHULMAN K A, NIELSEN P K Jr, PATEL K. AI alone will not reduce the administrative burden of health care[J]. JAMA, 2023, 330(22): 2159-2160.
[71] MELLO M M, ROSE S. Denial: artificial intelligence tools and health insurance coverage decisions[J]. JAMA health forum, 2024, 5(3): e240622.
[72] HARRER S. Attention is not all you need: the complicated case of ethically using large language models in healthcare and medicine[J]. eBioMedicine, 2023, 90: 104512.
[73] MINSSEN T, VAYENA E, GLENN COHEN I. The challenges for regulating medical use of ChatGPT and other large language models[J]. JAMA, 2023, 330(4): 315-316.
[74] SEZGIN E, SIRRIANNI J, LINWOOD S L. Operationalizing and implementing pretrained, large artificial intelligence linguistic models in the US health care system: outlook of generative pretrained transformer 3 (GPT-3) as a service model[J]. JMIR medical informatics, 2022, 10(2): e32875.
[75] SUN L, HUANG Y, WANG H, et al. Trustllm: trustworthiness in large language models[EB/OL]. (2024-01-10)[2025-08-29]. https://arxiv.org/abs/2401.05561.
[76] CABRAL S, RESTREPO D, KANJEE Z, et al. Clinical reasoning of a generative artificial intelligence model compared with physicians[J]. JAMA internal medicine, 2024, 184(5): 581-583.
[77] BHAYANA R. Chatbots and large language models in radiology: a practical primer for clinical and research applications[J]. Radiology, 2024, 310(1): e232756.
[78] LI Hanzhou, MOON J T, PURKAYASTHA S, et al. Ethics of large language models in medicine and medical research[J]. The lancet digital health, 2023, 5(6): e333-e335.
- 相似文献/References:
-
[1]李德毅.网络时代人工智能研究与发展[J].智能系统学报,2009,4(1):1.
LI De-yi.AI research and development in the network age[J].CAAI Transactions on Intelligent Systems,2009,4():1.
[2]赵克勤.二元联系数A+Bi的理论基础与基本算法及在人工智能中的应用[J].智能系统学报,2008,3(6):476.
ZHAO Ke-qin.The theoretical basis and basic algorithm of binary connection A+Bi and its application in AI[J].CAAI Transactions on Intelligent Systems,2008,3():476.
[3]徐玉如,庞永杰,甘?? 永,等.智能水下机器人技术展望[J].智能系统学报,2006,1(1):9.
XU Yu-ru,PANG Yong-jie,GAN Yong,et al.AUV—state-of-the-art and prospect[J].CAAI Transactions on Intelligent Systems,2006,1():9.
[4]王志良.人工心理与人工情感[J].智能系统学报,2006,1(1):38.
WANG Zhi-liang.Artificial psychology and artificial emotion[J].CAAI Transactions on Intelligent Systems,2006,1():38.
[5]赵克勤.集对分析的不确定性系统理论在AI中的应用[J].智能系统学报,2006,1(2):16.
ZHAO Ke-qin.The application of uncertainty systems theory of set pair analysis (SPU)in the artificial intelligence[J].CAAI Transactions on Intelligent Systems,2006,1():16.
[6]秦裕林,朱新民,朱? 丹.Herbert Simon在最后几年里的两个研究方向[J].智能系统学报,2006,1(2):11.
QIN Yu-lin,ZHU Xin-min,ZHU Dan.Herbert Simons two research directions in his lost years[J].CAAI Transactions on Intelligent Systems,2006,1():11.
[7]谷文祥,李 丽,李丹丹.规划识别的研究及其应用[J].智能系统学报,2007,2(1):1.
GU Wen-xiang,LI Li,LI Dan-dan.Research and application of plan recognition[J].CAAI Transactions on Intelligent Systems,2007,2():1.
[8]杨春燕,蔡 文.可拓信息-知识-智能形式化体系研究[J].智能系统学报,2007,2(3):8.
YANG Chun-yan,CAI Wen.A formalized system of extension information-knowledge-intelligence[J].CAAI Transactions on Intelligent Systems,2007,2():8.
[9]赵克勤.SPA的同异反系统理论在人工智能研究中的应用[J].智能系统学报,2007,2(5):20.
ZHAO Ke-qin.The application of SPAbased identicaldiscrepancycontrary system theory in artificial intelligence research[J].CAAI Transactions on Intelligent Systems,2007,2():20.
[10]王志良,杨?? 溢,杨?? 扬,等.一种周期时变马尔可夫室内位置预测模型[J].智能系统学报,2009,4(6):521.[doi:10.3969/j.issn.1673-4785.2009.06.009]
WANG Zhi-liang,YANG Yi,YANG Yang,et al.A periodic time-varying Markov model for indoor location prediction[J].CAAI Transactions on Intelligent Systems,2009,4():521.[doi:10.3969/j.issn.1673-4785.2009.06.009]
备注/Memo
收稿日期:2024-10-15。
基金项目:国家自然科学基金项目（92259104）.
作者简介:王璐，博士研究生，主要研究方向为医学数据分析、自然语言处理。发表学术论文论文7篇。 E-mail：luwang@sj-hospital.org。;丁慕菲，硕士研究生，主要研究方向为医学图像处理。 E-mail：dingmuou@163.com。;宋江典，副教授，博士，中国计算机学会数字医学分会执行委员，主要研究方向为医学图像处理与人工智能，主持国家自然科学基金项目2项。发表学术论文34篇。 E-mail：jdsong@cmu.edu.cn。
通讯作者:宋江典. E-mail：jdsong@cmu.edu.cn
更新日期/Last Update:
1900-01-01