名中医不寐病“AI智能人”数据

用于构建、训练并持续优化能够高度复现、传承并辅助扩展张永华名中医临床诊疗思想与经验的大模型或智能体。主要适用于失眠、焦虑障碍、抑郁症等心身疾病，可以应用于对中医智能化或诊疗研究的人工智能技术公司、中医药管理机构、卫生机构等。本数据基于杭州市中医院信息系统中张永华名中医团队2024年门诊病例构建，依据诊断关键字筛选睡眠障碍相关样本。数据处理遵循脱敏分段和质量可控原则经K-匿名隐私保护模型处理；诊断信息经结构化整合生成中医诊断、中医证型及西医诊断字段，处方信息自动聚合为草药方与西处方。以核心字段完整性、诊断符合性和文本可读性进行质量评价。经专家评估后对样本实施赋分筛选：诊断结构完整（具备标准化中医诊断或西医诊断名称，中医诊断需要配有中医证型，多条诊断通过结构化字段完整呈现，中西医诊断缺失或无法解析的不得分。占比40分）、处方可解析（草药与西药字段需包含药物名称及组成信息，可被系统解析出有效条目。中西医处方同时缺失或文本无效不得分。占比40分）、主诉和病史文本有效（主诉与现病史、既往史字段须具备明确临床语义，文本长度及关键词覆盖满足基本问诊表达，无歧义内容，排除空值或无效描述。占比20分）三项评分加权，≥80 分定义为高质量样本；60–79 分为中质量样本；＜60 分样本为备选样本，按此排列进行择优选择并经过专家人工筛选，张永华名中医工作室医生对数据进行医学逻辑和准确性校验，确保数据合法合规、结构标准、语义清晰，可稳定支撑名中医知识沉淀与应用服务落地。

This dataset is developed for building, training and continuously optimizing large language models (LLMs) or AI Agents that can highly replicate, inherit and assist in expanding the clinical diagnosis and treatment thoughts and experience of renowned TCM physician Zhang Yonghua. It is primarily targeted at psychosomatic diseases including insomnia, anxiety disorders and depression, and is applicable to AI technology companies engaged in TCM intelligentization or diagnosis and treatment research, TCM management authorities, medical and health institutions, etc.

This dataset is constructed based on 2024 outpatient cases of the renowned TCM physician Zhang Yonghua's team retrieved from the information system of Hangzhou Hospital of Traditional Chinese Medicine, with sleep disorder-related samples screened using diagnostic keywords.

Data processing follows the principles of desensitization, segmentation and quality control, and is processed through the K-anonymity privacy protection model. Diagnostic information is structurally integrated to generate three fields: TCM diagnosis, TCM syndrome type and Western medicine diagnosis; prescription information is automatically aggregated into herbal prescriptions and Western medicine prescriptions.

Quality assessment is conducted based on three dimensions: core field completeness, diagnostic compliance and text readability. Samples are scored and screened following expert review, with three weighted scoring items as follows:
1. Complete diagnostic structure (40 points): Samples must have standardized TCM diagnosis or Western medicine diagnosis names; TCM diagnoses must be paired with corresponding TCM syndrome types; multiple diagnoses must be fully presented via structured fields. Samples with missing or unresolvable Chinese and Western medicine diagnoses will receive zero points.
2. Parsable prescriptions (40 points): Both herbal and Western medicine prescription fields must include drug names and composition information, and can be parsed into valid entries by the system. Samples with both Chinese and Western medicine prescriptions missing or with invalid text will receive zero points.
3. Valid chief complaint and medical history text (20 points): The chief complaint, current medical history and past medical history fields must contain clear clinical semantics, meet basic consultation expression requirements in terms of text length and keyword coverage, have no ambiguous content, and exclude null values or invalid descriptions.

Samples with a total score of ≥80 are defined as high-quality samples; those with 60–79 points are medium-quality samples; samples with a score of <60 are classified as alternative samples. Samples are selected according to this ranking, followed by expert manual screening. Physicians from the renowned TCM physician Zhang Yonghua's studio conduct medical logic and accuracy verification on the data, ensuring that the data is legally compliant, structurally standardized, semantically clear, and can stably support the precipitation and application of renowned TCM physician knowledge for service implementation.

搜集汇总

背景与挑战

背景概述

该数据集是杭州市中医院登记的名中医不寐病“AI智能人”数据，包含500条高质量门诊病例样本，格式为xlsx，用于构建和优化能复现张永华名中医诊疗经验的大模型或智能体。数据基于2024年病例构建，经过脱敏、结构化处理和质量评估，适用于失眠、焦虑障碍等心身疾病的中医智能化研究和应用。