视觉语言模型的基准测试用于紧急和重症监护机构的诊断
Christoph F Kurz1, Tatiana Merzhevich2, Bjoern M Eskofier2,3
1Novartis Pharma GmbH, Nuremberg, Germany. christoph.kurz@novartis.com.
NPJ digital medicine
|July 10, 2025
概括
视觉语言模型 (VLMs) 在急性护理环境中显示出有限的诊断准确性. 需要专门的培训来改进用于紧急和重症监护应用的开源VLM,因为GPT-4o展示了卓越的性能.
科学领域:
- 人工智能在医学中的应用
- 医疗信息学 医疗信息学
- 临床决策支持系统 临床决策支持系统
背景情况:
- 视觉语言模型 (VLM) 是新兴的人工智能工具,在医疗保健领域具有潜力.
- 它们在急症和重症监护病房等高急症临床环境中的应用尚未得到充分理解.
- 目前的开源VLM可能无法满足急性护理的诊断需求.
研究的目的:
- 在急性护理场景中评估小型开源VLMs的诊断性能.
- 将开源VLM的功能与最先进的模型 (GPT-4o) 进行比较.
- 确定紧急医疗的VLM的局限性和改进领域.
主要方法:
- 使用多式联络数据集,包括诊断问题,医疗图像和临床背景.
- 基准测试了几种小型的,开源的视觉语言模型.
- 与先进的GPT-4o模型进行性能比较.
主要成果:
- 开源的VLM实现了最大的诊断准确率为40.4%.
- GPT-4o显著优于开源模型的性能,达到68.1%的准确性.
- 目前的开源VLM与急症护理的先进模型之间存在着相当大的性能差距.
结论:
- 目前,开源的VLM在关键急性护理应用中显示出不足的诊断准确性.
- 开源VLM需要在专业培训和模型优化方面取得重大进展.
- GPT-4o显示出有希望的结果,但需要进一步验证,以便在紧急和重症监护中进行临床部署.
相关概念视频
01:14SBAR II: Application of SBAR
4.8K
SBAR is an effective communication tool used by healthcare professionals to communicate patient information accurately. SBAR stands for Situation, Background, Assessment, and Recommendation. For a better understanding, an example is given below.
SBAR Report from a Nurse to a Health Care Provider
S: "Hello, Dr. Smith. This is Jane, RN, from the Med Surg unit. I am calling to tell you about Ms. White in Room 210, who is experiencing increased pain and redness at her incision site. Her recent...
SBAR Report from a Nurse to a Health Care Provider
S: "Hello, Dr. Smith. This is Jane, RN, from the Med Surg unit. I am calling to tell you about Ms. White in Room 210, who is experiencing increased pain and redness at her incision site. Her recent...
4.8K
01:22Cardiopulmonary Resuscitation II: ACLS Airway Management
109
Airway management is a key skill in emergency and critical care settings, as maintaining a clear airway is essential for adequate oxygenation and ventilation.Head Tilt-Chin Lift TechniqueThe head tilt-chin lift maneuver is an essential technique primarily used in patients without suspected cervical spine injuries. To perform this maneuver, one hand is placed on the patient’s forehead, and gentle pressure is applied backward to tilt the head. The fingertips of the other hand are positioned...
109