Friedlin Jeff, McDonald Clement J
Regenstrief Institute, Inc, Indiana University School of Medicine, Indianapolis, IN, USA.
AMIA Annu Symp Proc. 2006;2006:925.
We developed a rule-based natural language processing (NLP) system for extracting and coding clinical data from free text reports. We studied the systems ability to accurately extract and code family history data from hospital admission notes. The system searches the family history for 12 diseases (and relative degree). It achieved a sensitivity of .96 and a PPV of .97 for disease extraction, and .96 and .93 respectively for relative categorization.
我们开发了一个基于规则的自然语言处理(NLP)系统,用于从自由文本报告中提取和编码临床数据。我们研究了该系统从医院入院记录中准确提取和编码家族史数据的能力。该系统在家族史中搜索12种疾病(以及亲属关系程度)。在疾病提取方面,其灵敏度为0.96,阳性预测值为0.97;在亲属关系分类方面,灵敏度和阳性预测值分别为0.96和0.93。