基于条件随机域的禤国维名老中医医案挖掘分析

Text Mining and Analysis of Medical Records from Famous Chinese Medicine Expert XUAN Guo-wei Based on Conditional Random Fields

目的:基于条件随机域模型对禤国维名老中医临床医案进行挖掘分析。方法:基于条件随机域设计一种新的临床医案文本挖掘信息模型,利用该模型对国医大师禤国维医案进行临床术语信息的自动识别提取,并使用关联规则算法进行数据挖掘分析。结果:基于条件随机域构建的文本信息挖掘模型对不同临床术语类型,不同的病种识别效果良好,其中在不同术语类型的识别中对"中医诊断术语"的F-测度值最高,达到89.87%,在不同病种的识别中对"湿疹"的F-测度值最高,达到83.77%。对其中禤国维名老中医治疗荨麻疹的经验进行挖掘,发现在治法上多采取益气固表、祛风止痒、除湿解毒;常用以治疗的二联药对有黄芪-白术、黄芪-防风、白术-防风等,其中白术-防风的置信度(1.00)和支持度(0.83)最高;常用的三联药物有黄芪-紫苏叶-白术等,常见的药物症状关联有地黄-风团、防风-瘙痒等,其中地黄-风团的置信度(1.00)和支持度(0.97)最高。结论:该研究基于条件随机域构建了一种新的文本挖掘信息模型,该模型有利于高效整理和挖掘名老中医临床医案中的经验与学术思想,对名老中医的学术思想和经验传承具有重要意义。

Objective： To explore and analyze the clinical medical records from famous Chinese medicine expert XUAN Guo-wei based on the conditional random field. Method： A new text mining model for the medical records were designed based on conditional random field. Then this model was used to automatically recognize and extract the clinical terms information in the medical records from famous Chinese medicine expert XUAN Guo-wei,and association rules algorithm was used for data mining analysis. Result： The text information mining model which was constructed based on conditional random fields,can effectively identify different clinical terms and different disease types; in the results of recognizing different types of terms,‘traditional Chinese medicine（ TCM）diagnosis terms＇ had the highest F-measure value,reaching 89. 87%; in the results of recognizing different disease types,‘eczema ＇ had the highest F-measure value,reaching 83. 77%. Based on XUAN Guo-wei＇s clinical experience in the treatment of urticaria： professor Xuan often used the methods of tonifying Qi and strengthening exterior,dispelling wind and arresting itching,as well as removing dampness and detoxifying for urticaria patients;his commonly used herbal pairs included Astragali Radix-Atractylodis Macrocephalae Rhizoma,Astragali RadixSaposhnikoviae Radix,Atractylodis Macrocephalae Rhizoma-Saposhnikoviae Radix,etc,and the confidence degree and support degree of Atractylodis Macrocephalae Rhizoma-Saposhnikoviae Radix were highest, respectively reaching 1. 00 and 0. 83; his commonly used triple herbal combinations included Astragali Radix-Perillae FoliumAtractylodis Macrocephalae Rhizoma,etc. Common TCM-symptom association pairs included Rehmanniae Radixwheals,Saposhnikoviae Radix-itching,etc,and the confidence degree and support degree of Rehmanniae Radixwheals were highest,respectively reaching 1. 00 and 0. 97. Conclusion： A new text information mining model was constructed in this study based on conditional random fields,which was conducive to efficiently excavate and succeed the clinical medical experience and academic thoughts of famous old Chinese Medicine experts.