ACI-BENCH

ACI-BENCH是一个创新的环境临床智能数据集，专为基准测试自动访问笔记生成而设计。该数据集是目前最大的，包含207条记录，由微软健康与生命科学人工智能团队和华盛顿大学生物医学与健康信息学团队共同创建。ACI-BENCH旨在模拟医生-患者对话中的三种模型辅助临床笔记生成变体，包括虚拟助手交互、无约束的抄写员讨论以及医生与患者之间的自然对话。数据集的创建过程涉及医学专家，包括医生、助理医师、医学抄写员和临床信息学家，基于实际遭遇的经验和研究。ACI-BENCH的应用领域主要集中在通过自动摘要技术辅助医生记录临床笔记，以提高医疗记录的效率和准确性，同时减轻医生的文档负担。

ACI-BENCH is an innovative environmental clinical intelligence dataset designed for benchmarking automatic visit note generation. It is currently the largest of its kind, containing 207 records, co-developed by the Microsoft Health & Life Sciences AI Team and the University of Washington's Biomedical and Health Informatics team. ACI-BENCH aims to simulate three model-assisted clinical note generation variants in physician-patient conversations, including virtual assistant interactions, unconstrained scribe discussions, and natural dialogues between physicians and patients. The dataset was developed with involvement from medical professionals including physicians, physician assistants, medical scribes, and clinical informaticians, based on real-world clinical encounter experiences and relevant research. The primary application scope of ACI-BENCH focuses on assisting physicians in documenting clinical notes through automatic summarization technologies, to improve the efficiency and accuracy of medical records while alleviating physicians' documentation burden.