基于SMOTE算法构建子宫内膜异位症术后复发的风险模型  

Constructing a Risk Model for Postoperative Recurrence of Endometriosis Based on the SMOTE Algorithm

作　　者：董纪秀 梁惠霞 胡燕[1] 吕静雯 许珠婷 石燕梅 林淑媛 DONG Jixiu;LIANG Huixia;HU Yan;L Jingwen;XU Zhuting;SHI Yanmei;LIN Shuyuan(Department of Obstetrics,The 910th Hospital of the Chinese People's Liberation Army Joint Logistic Support Force,Quanzhou 362000,China)

机构地区：[1]中国人民解放军联勤保障部队第910医院妇产科,泉州362000

出　　处：《福建医科大学学报》2024年第5期331-336,共6页Journal of Fujian Medical University

摘　　要：目的探究基于合成少数类过采样技术(SMOTE)算法构建子宫内膜异位症(EMs)术后复发的风险模型。方法收集2017年1月—2023年3月行腹腔镜保守性手术的EMs患者148例,根据术后复发情况分为复发组(30例)和未复发组(118例)。回顾性分析受试者的资料,采用单因素及logistic回归分析筛选EMs患者术后复发的危险因素,通过SMOTE算法重建影响因素原始数据集,得出风险预警模型并验证其预测效能。结果148例中,术后复发30例(20.27%)。2组的体质量指数(BMI)、病程、术前并发症、既往妇科手术史、术前产次、囊肿直径、多房多囊等比较,差别均无统计学意义(P>0.05);年龄、术前痛经、病变侧别、rAFS分期、术后妊娠、术后辅助药物治疗等比较,差别均有统计学意义(P<0.05)。logistic回归分析结果显示,年龄较小、术前痛经史、双侧病变、术后妊娠、术后未采用辅助药物治疗是EMs患者术后复发的危险因素(P<0.05);得到原始预警模型和基于SMOTE算法的预警模型,经H-L检验显示,模型拟合度均良好,受试者工作特征(ROC)曲线下面积(AUC)分别为0.854和0.869,DeLong P值为0.048(P<0.05)。结论基于年龄较小、术前痛经史、双侧病变、术后妊娠、术后未采用辅助药物治疗原始数据和SMOTE算法建立的预警模型均具有较高的预测性,医护人员可据此进行有效干预,以预判EMs患者的术后复发情况。Objective To investigate the construction of a risk model for postoperative recurrence of endometriosis(EMs)based on the synthetic minority over-sampling technique(SMOTE)algorithm.Methods A total of 148 patients with EMs who underwent conservative laparoscopic conservative surgery from January 2017 to March 2023 were retrospectively collected as observation subjects,and the patients with EMs were divided into a recurrence group(30 cases)and a non-recurrence group(118 cases)according to their postoperative recurrence.The data of the subjects were retrospectively analyzed,and the risk factors for postoperative recurrence of EMs patients were screened by using univariate and logistic regression analyses,and then the original dataset of the influencing factors was reconstructed by the SMOTE algorithm to derive a risk warning model and validate its predictive efficacy.Results There were 30 cases of postoperative recurrence in 148 cases of EMs,with an incidence rate of 20.27%;there was no statistically significant difference in the comparisons of body mass index(BMI),duration of disease,preoperative complications,history of previous gynecological surgeries,preoperative deliveries,diameter of cysts,and polycysticity between the two groups(P>0.05);there was a statistically significant difference in age,preoperative dysmenorrhea,side of lesion,rAFS stage,postoperative pregnancy,and postoperative adjuvant medication(P<0.05);logistic regression analysis showed that younger age,preoperative history of dysmenorrhea,bilateral lesions,postoperative pregnancy,and postoperative adjuvant medication not used were the risk factors for postoperative recurrence in patients with EMs(P<0.05);the original early warning model was obtained,the early warning model based on the SMOTE algorithm,and the results of H-L test showed that the model fit was good.Analyzing the ROC curves of the early warning model resulted in the AUC of 0.854 and 0.869,with a DeLong P-value of 0.048(P<0.05).Conclusion Both the early warning models built on the origi

关 键 词：SMOTE算法 子宫内膜异位症 术后复发 

分 类 号：R713[医药卫生—妇产科学]