智慧医检4.0时代来临：域见医言大模型助力医检服务更精准
随着智慧医检进入4.0时代，医学检验服务的全场景智能化进程不断加速。如何让医检服务更精准、更方便？8月23日，中华医学会第十八次检验医学学术会议在浙江杭州举行，会上发布了第三方医检行业首个医检大模型——域见医言大模型，并同步上线智能体应用“小域医”。该模型的上线将推动医检服务从检前项目查询到检中生产作业，再到检后报告解读的全流程智能化，减轻医生负担，优化患者就诊体验。
医学检测是医学诊疗过程中的关键一环，70%的临床决策信息来自医学实验室结果。然而，随着精准医学的发展，检验医学面临诸多挑战：传统操作流程繁琐、报告判读和分析工作量大、生产效率有待提升；复杂多变的临床场景要求临床医生和检验医师具备更快整合、处理检测信息的能力；检验医师队伍建设面临培养周期长、交叉学科难度大等问题。
“在当前的精准医疗时代下，检验人需要有临床思维，也要有数据思维。”浙江省肿瘤医院检验科主任张毅敏表示，医检大模型可以将临床、检验和数据融合得更紧密，有利于医生出具更为精准的检查报告，同时惠及患者。
重庆医科大学科研处副处长、西部智慧检验与数字医疗协同创新中心主任李小松强调，医检大模型的开发过程中，尤其要注重临床数据的精准匹配和安全性，只有注重海量数据的安全、可靠、合规，加上多学科、跨学科的人才，才有可能做出更为实用的医检大模型。
金域医学副总裁兼数字化管理中心总经理李映华认为，一个真正满足临床需求的医检行业大模型，需要具备多模型整合、多场景赋能、多模态与多组学数据处理、多元知识充分融合、基于全病程信息多轮互动等能力，不仅仅具备知识问答、智能化检测结果整合的功能，还具备医学领域的专业分析、推理能力。
此次发布的域见医言大模型，基于23000余家医疗机构的服务经验，在通用语料基础上注入了超20亿Token(数据单元)医检语料，经过近两年的开发训练而成。目前已有超2万名企业专业技术人员、临床专家、检验医师参与测试。
据了解，该模型有望支持整合图像、语音、文本等多模态，基因、蛋白、病理等多组学，以及项目推荐、实验室检测、报告生成等多场景的信息，能处理复杂任务，帮助临床医生、检验医师在多轮交互中作出科学决策，并推动医学实验室检测环节智能生产、检测结果智能判读，实现医学检测的检前、检中、检后环节无缝衔接。
“和其他行业大模型不一样的是，‘域见医言’不依赖特定大模型底座，而是能适应各类通用多模态大模型，并支持衔接形态学、病理、基因等专业领域大模型。同时，其也可在未来迭代升级，不断学习新的知识和信息，为智慧医检4.0时代提供更强大的智能引擎。”域见医言大模型研发团队表示。
域见医言大模型的发布，标志着智慧医检4.0时代正式来临。未来，随着人工智能技术的不断发展，医检服务将更加精准、高效、便捷，为患者提供更优质的医疗服务。
英语如下：
Smart Medical Examination 4.0: Precision Service, a Promising Future!
Keywords: Smart Medical Examination, Precision Service, Intelligence
TheEra of Smart Medical Examination 4.0 Arrives: Domain-Specific Medical Language Model Powers More Precise Medical Examination Services
As smart medical examination enters the4.0 era, the intelligent transformation of medical examination services across all scenarios is accelerating. How can we make medical examination services more precise and convenient? On August23rd, the 18th National Congress of Medical Laboratory Science of the Chinese Medical Association was held in Hangzhou, Zhejiang Province, where the first medical examination large language model in the third-party medical examination industry, Domain-Specific Medical Language Model, was released, along with the launch of the intelligent agent application “Xiao Yu Yi.” The launch of this model will drive the intelligent transformation of medical examination services, from pre-examination project inquiries to in-examinationproduction operations, and finally to post-examination report interpretation, alleviating the burden on doctors and optimizing the patient’s medical experience.
Medical examination is a crucial link in the medical diagnosis and treatment process, with 70% of clinical decision information coming from medical laboratory results. However, with the development of precisionmedicine, medical examination faces numerous challenges: traditional operational processes are cumbersome, report interpretation and analysis workload is large, and production efficiency needs improvement; complex and ever-changing clinical scenarios require clinicians and laboratory physicians to have the ability to integrate and process test information more quickly; the development of the laboratory physician workforce faces challenges such aslong training cycles and difficulties in cross-disciplinary training.
“In the current era of precision medicine, laboratory professionals need to have both clinical thinking and data thinking,” said Zhang Yimin, Director of the Department of Laboratory Medicine at Zhejiang Provincial Cancer Hospital. “Medical examination large language models can integrate clinical, laboratory, anddata more closely, which is beneficial for doctors to issue more precise examination reports and benefit patients.”
Li Xiaosong, Deputy Director of the Scientific Research Office of Chongqing Medical University and Director of the Western Collaborative Innovation Center for Smart Medical Examination and Digital Medicine, emphasized that in the development of medical examination large language models, particular attention should be paid to the precise matching and security of clinical data. Only by focusing on the safety, reliability, and compliance of massive data, combined with multidisciplinary and cross-disciplinary talents, can we create more practical medical examination large language models.
Li Yinghua, Vice President and General Manager of theDigital Management Center of KingMed Diagnostics, believes that a truly clinical-need-driven medical examination industry large language model needs to possess capabilities such as multi-model integration, multi-scenario empowerment, multi-modal and multi-omics data processing, multi-source knowledge integration, and multi-round interaction based on full-course information. It should not only have knowledge question answering and intelligent integration of test results functions but also possess professional analysis and reasoning capabilities in the medical field.
The Domain-Specific Medical Language Model released this time is based on the service experience of over 23,000 medical institutions. It was developed andtrained over nearly two years, incorporating over 2 billion tokens (data units) of medical examination materials on top of general language materials. Currently, over 20,000 enterprise professionals, clinical experts, and laboratory physicians have participated in testing.
It is understood that the model is expected to support the integrationof multi-modal information such as images, voice, and text, multi-omics data such as genes, proteins, and pathology, and multi-scenario information such as project recommendations, laboratory testing, and report generation. It can handle complex tasks, helping clinicians and laboratory physicians make scientific decisions through multi-round interactions andpromoting intelligent production of medical laboratory testing processes and intelligent interpretation of test results, achieving seamless connection between pre-examination, in-examination, and post-examination stages of medical examination.
“Unlike other industry large language models, ‘Domain-Specific Medical Language’ does not rely on a specific large model base but can adaptto various general multi-modal large models and support the connection of large models in specialized fields such as morphology, pathology, and genetics. At the same time, it can also be iteratively upgraded in the future, continuously learning new knowledge and information, providing a more powerful intelligent engine for the smart medical examination 4.0 era,” said the Domain-Specific Medical Language Model research and development team.
The release of the Domain-Specific Medical Language Model marks the official arrival of the smart medical examination 4.0 era. In the future, with the continuous development of artificial intelligence technology, medical examination services will become more precise, efficient, and convenient, providing patients with higher-quality medical services.
【来源】http://www.chinanews.com/sh/2024/08-23/10273722.shtml
Views: 1