人工智能自然语言处理诊断肾癌T分期的研究

Research on Artificial Intelligence Natural Language Processing in Diagnosing T-stage of Renal Carcinoma

目的 探讨人工智能自然语言处理方法诊断肾癌T分期的可行性和准确性。方法 收集2018年1月~2020年1月国家癌症中心、国家肿瘤临床医学研究中心、中国医学科学院/北京协和医学院肿瘤医院200例肾癌患者作为训练组,选取2015年1月~2017年12月性别、年龄、病理分期匹配的200例患者作为测试组,使用基于规则匹配和条件随机场两种人工智能自然语言处理方法对病理文本数据进行提取分析。结果 基于规则匹配和条件随机场两种方法在测试组的正确率分别为99.0%和95.5%。测试组的方法性能评估中,规则匹配方法的准确率为99.0%,召回率为99.0%,F1-分数为99.0%。条件随机场方法的准确率为97.1%,召回率为95.5%,F1-分数为96.3%。结论 人工智能通过自然语言处理方法自动诊断肾癌T分期可行,且基于规则匹配的算法准确性较高,这些研究结果有待于多中心、大样本数据进一步验证。

Objective To explore the feasibility and accuracy of artificial intelligence natural language processing in diagnosing T-stage of renal carcinoma.Methods A total of 200 patients with renal carcinoma from National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital,Chinese Academy of Medical Sciences and Peking Union Medical College from January 2018 to January 2020 were selected as the training group,and 200 patients with matched gender,age and pathological stage between January 2015 to December 2017 were selected as the test group.Two artificial intelligence natural language processing methods including rule based template matching and conditional random fields were used to extract and analyze pathological text data.Results The accuracy rates of rule-based template matching and conditional random fields were 99.0%and 95.5%respectively.In the method performance evaluation of the test group,the accuracy rate of the rule-based template matching was 99.0%,the recall rate was 99.0%,and the F 1-score was 99.0%.The accuracy rate of the conditional random fields was 97.1%,the recall rate was 95.5%,and an F 1-score was 96.3%.Conclusion It is feasible for artificial intelligence to automatically diagnose the T-stage of renal carcinoma by natural language processing,and the algorithm of rule-based template matching had high accuracy.These research results still need to be further verified by multi-center and large sample data.