Abstract:

A preliminary study of the significance of artificial intelligence in the clinical practice of breast cancer postoperation        DUAN Ming-yue*，YE Yu-qin，ZHANG Le，et al.*Department of Breast Surgery，China-Japan Union Hospital of Jilin University，Changchun 130033，China
Corresponding authors：WANG Ke-ren，E-mail：dearfad@sina.com；YE Yu-qin，E-mail：dearlea@live.cn
Abstract    Objective    To explore the guiding significance of artificial intelligence for clinical practice of breast cancer after operation，and to evaluate the concordance of the Watson for oncology（WFO） with the breast cancer treatment guidelines of NCCN and Chinese Anti-Cancer Association. Methods    A total of 100 patients who were diagnosed as breast cancer after accepting surgical operations from January 2016 to December 2017 in China-Japan Union Hospital of Jilin University were enrolled. The expressions of ER，PR，Her-2 and Ki-67 protein in cancer tissues after operation were detected by using immunohistochemitry，and inputed the clinical cases information of these 100 patients into the WFO system. Finally，the treatment recommendations were compared with the breast cancer treatment guidelinesof NCCN and the China Anticancer Association. The Kappa test was used to determine the concordance. Results    There were 64 cases （64%） of all 100 patients with the same treatment plans，the clinical stage Ⅱ（42/44cases，95.5%）to Ⅲ（15/16cases，93.6%） and triple negative breast cancer（7/8cases，87.5%） treatment showed a higher concordance （Kappa value ≥ 0.75）；The age≥41 years old，Luminal A of the immunophenotype and pathological grade Ⅲ also had a certain concordance，and the concordance of 41-55 years old were average，but it is not ideal of the cases that older than 55 years old or the pathological grade Ⅲ；On the contrary，the treatment recommendations of clinical stage Ⅰ（7/40  cases，17.5%），especially for both the lymph node metastasis negative were very inconsistency（Kappa value ＜0）.But there was no statistical significance in the cases of the age ≤ 41 years old，pathological gradeⅠto Ⅱ，Luminal B and HER-2 type of immunophenotype. Conclusion    WFO in the patients of clinical stageⅡ to Ⅲ and triple negative breast cancer（P＜0.05）has a certain reference value in the afore-mentioned situations；And WFO could be considering in the cases of age ≥ 41 years of age or older，Luminal A of immunophenotype and pathological grade Ⅲ；In contrast，for clinical stage Ⅰ of early breast cancer patients，WFO is not recommended. WFO could offer some reasonable treatment recommendations in partial patients with artificial intelligent computer system，and it may provide clinical decision-support for clinicians，and it has some guiding significance in breast cancer postoperative clinical practice，clinical learning and research.

Key words: breast cancer postoperative, artificial intelligence, Watson for oncology, clinical decision-support, concordance

摘要：

目的    探讨人工智能对乳腺癌术后临床实践的指导意义，评估 Watson肿瘤解决方案（WFO）与美国国家综合癌症网络（NCCN）和中国抗癌协会乳腺癌治疗指南的一致性。方法    回顾性分析吉林大学中日联谊医院2016年1月至2017年12月期间实施手术后确诊为乳腺癌的100例病人资料，术后癌组织标本均采用免疫组化方法检测癌组织中雌激素受体（ER）、孕激素受体（PR）、人类表皮生长因子受体-2（HER-2）及Ki-67的表达水平，将该100例病人的相关临床资料录入到WFO系统中，并将其推荐治疗方案与NCCN及中国抗癌协会乳腺癌治疗指南进行对比，采用Kappa检验分析其一致性。结果    100例病人中共有64例（64%）治疗方案一致，其中临床分期为Ⅱ期（42/44，95.5%）～Ⅲ期（15/16，93.6%）及免疫组化分型为三阴性乳腺癌（7/8，87.5%）的治疗方案表现出较高的一致性（Kappa值≥0.75）；年龄≥41岁、免疫组化分型为Luminal A型及病理学分级为Ⅲ级的治疗方案也都具有一定的一致性，其中年龄41~55岁的病例的治疗方案一致性一般，而年龄＞55岁级及病理学分级为Ⅲ级的治疗方案一致性不理想；在年轻早期乳腺癌病人中，临床分期Ⅰ期（7/40，17.5%）、淋巴结转移癌为阴性的治疗方案很不一致（Kappa值＜0）；而年龄＜41岁、病理学分级为Ⅰ～Ⅱ级、免疫组化分型为Luminal B型及HER-2型的病例的治疗方案无统计学意义。结论    WFO在临床分期Ⅱ～Ⅲ期及免疫分型为三阴性乳腺癌（P＜0.05）病人中具有较高的参考价值；在年龄≥41岁、分子分型为Luminal A型、病理学分级为Ⅲ级的病人可考虑使用；对临床分期Ⅰ期的早期乳腺癌病人不推荐使用。WFO对部分病人可给出合理的治疗建议，能为临床医生提供临床决策支持，在乳腺癌术后临床实践方面及临床学习和研究中具有一定的指导意义。

关键词: 乳腺癌术后, 人工智能, 肿瘤解决方案, 临床决策, 一致性