目的 藏医证型辨识的准确性是有效地指导临床实践的重要因素，对进一步确定诊疗标准起着很大作用。针对传统藏医尿诊依赖临床医生的丰富经验并存在主观差异的问题，开展基于尿诊医案信息的证型分类的客观化研究，基于类属属性构建证型多标签分类模型，实现藏医尿诊辅助证型辨识。方法 收集915例青海省藏医院和青海久美藏医院住院患者的尿诊医案信息，建立藏医尿诊医案数据库，根据藏医理论对尿诊医案信息进行规范化和编码处理，以不同证型标签的正负样本中心为基础基于类属属性构建分类预测模型。结果 所构建模型根据患者医案信息进行多标签分类，获得证型辨识结果（汉明损失0.2111、排序损失0.3055、0-1损失0.3579、覆盖率0.7066、平均精准度81.71%。结论 模型在藏医尿诊证型辨识问题中表现出应用价值，为藏医尿诊的计算机辅助诊疗提供新思路与方法。

Objective The syndrome differentiation of Tibetan medicine is highly important for doctors to effectively guide clinical practice, which plays a great role in further determining the diagnosis and treatment standards. Aiming at the problem that the traditional Tibetan urine diagnosis relies on the personal experience of clinicians, the research on syndrome differentiation based on the medical record information of urine diagnosis is carried out in this paper, and a multi-label syndrome classification model based on label-specific features is developed.Methods Firstly, the 915 hospitalized patients'medical records of Tibetan Medical Hospital of Qinghai Province and Tibetan Medical Hospital of Qinghai Jiumei are collected, standardized and encoded. Then, based on the positive and negative sample centers of different syndrome labels, the classification model is built.Results Finally, the results of syndrome classification are obtained: Hamming Loss 0.2111, Ranking loss 0.3055, One-Error 0.3579, Coverage 0.7066, Average Precision 81.71%.Conclusion The model shows its application value in the identification of Tibetan medicine urine diagnosis type, providing the new approach for the computer-aided diagnosis of urine diagnosis in Tibetan medicine.

TPxxx

国家科学技术部国家重点研发计划项目(2017YFC1703900):民族医药发掘整理与学术传承研究，负责人：张艺。国家自然科学基金委员会青年科学基金项目(81803851): 基于GIS和语义发掘的藏医高原病古籍文献治疗知识可视化发现研究，负责人：俞佳。