摘要
目的 构建和验证接受辅助生殖技术治疗患者的logistic 回归模型、反向传播（back propagation, BP）神经网络模型，并对其进行评价和比较，以期为提高临床妊娠率提供新策略。方法 回顾性分析2010年1月-2017年5月在湖南省某大型三甲医院医学生殖中心已经接受ART治疗且移植了胚胎的13207名患者的临床资料。通过单因素分析、多因素分析筛选出临床妊娠的影响因素，构建logistic回归模型和BP神经网络模型并验证，采用受试者工作特征（receiver operating characteristic, ROC）曲线评价模型方程。结果 本研究共12个预测变量进入最终的模型，分别为女方年龄、女方受教育程度、ART治疗次数、不孕年限、hCG日子宫内膜厚度、治疗方案、移植胚胎数量、移植胚胎质量、基础窦卵泡数、获卵总数、Gn总用量和Gn启动总天数。两模型间比较的结果显示，BP神经网络模型AUC更大（0.887），灵敏度（90.8%）和正确率（83.5%）更高，但特异度（49.6%）不理想；logistic回归模型特异度（74.7%）更高，但预测的灵敏度（66.1%）欠佳。结论 BP神经网络模型在预测临床妊娠方面有较高的应用价值，其与logistic回归模型结合应用，可互为补充，有助于实现临床妊娠的早期预测，并为进一步制定个体化治疗方案提供依据和参考，从而提高ART患者的临床妊娠率。
关键词： BP神经网络；logistic回归；辅助生殖技术；预测模型；临床妊娠
Abstract
Objective To construct and verify the logistic regression model and BP neural network model of assisted reproduction technology (ART) patients, to evaluate and compare the two models, and to provide a new strategy for improving clinical pregnancy rate. Methods From January 2010 to May 2017, the clinical data of 13207 patients who had received ART treatments and transplanted embryos in the Reproduction Center of Medicine of a large grade A hospital in Hunan Province were retrospective analyzed. The influencing factors of clinical pregnancy were analyzed by univariate analysis and multivariate analysis. Logistic regression model and BP neural network model were constructed and verified. The receiver operating characteristic (ROC) curve was used to evaluate the model equation. Results In this study, a total of 12 predictive variables were entered into the final model, female age, the female's level of education, No. of ART treatments, duration of infertility, endometrial thickness at the day of administration of hCG, therapeutic protocol, No. of embryos transplanted, quality of embryo transplanted, AFC, No. of oocytes retrieved, total dosage of Gn used and total duration of Gn used respectively. Compared with the two prediction models, BP neural network model had higher AUC (0.887), sensitivity (90.8%) and accuracy (83.5%), but it had lower specificity (49.6%). Logistic regression model had higher specificity (74.7%), but it had lower sensitivity (66.1%). Conclusion BP neural network model has high application value in predicting clinical pregnancy. It can be used in combination with logistic regression model to complement each other, contribute to achieve early prediction of clinical pregnancy, and provide basis and reference for further formulating individualized treatment plans, so as to improve the clinical pregnancy rate of ART patients.
Key words： BP neural network; Logistic regression; Assisted reproductive technology; Prediction model; Clinical pregnancy
参考文献 References
[1] 陆杰华, 林嘉琪. 中国人口新国情的特征、影响及应对方略——基于“七普”数据分析[J]. 中国特色社会主义研究, 2021, (03):57-67+2.
[2] Chambers GM, Dyer S, Zegers-Hochschild F, et al. International Committee for Monitoring Assisted Reproductive Technologies world report: assisted reproductive technology, 2014†[J]. Hum Reprod, 2021, 36(11): 2921-2934.
[3] 刘智慧, 李昆明, 董跃彦, 等. 辅助生殖反复种植失败患者心理弹性与生育生活质量关系的研究[J]. 解放军护理杂志, 2020, 37(10):35-38.
[4] 刘志强, 熊风, 张宏展, 等. IVF-ET妊娠结局预测模型的研究进展[J]. 生殖医学杂志, 2021, 30(05):695-700.
[5] Ratna MB, Bhattacharya S, Abdulrahim B, et al. A systematic review of the quality of clinical prediction models in in vitro fertilisation[J]. Hum Reprod, 2020, 35(1):100-116.
[6] Barnett-Itzhaki Z, Elbaz M, Butterman R, et al. Machine learning vs. classic statistics for the prediction of IVF outcomes[J]. J Assist Reprod Genet, 2020, 37(10):2405-2412.
[7] Curchoe CL, Malmsten J, Bormann C, et al. Predictive modeling in reproductive medicine: Where will the future of artificial intelligence research take us[J]. Fertil Steril, 2020, 114(5):934-940.
[8] Sfakianoudis K, Maziotis E, Grigoriadis S, et al. Reporting on the Value of Artificial Intelligence in Predicting the Optimal Embryo for Transfer: A Systematic Review including Data Synthesis[J]. Biomedicines, 2022, 10(3):697.
[9] Pergialiotis V, Pouliakis A, Parthenis C, et al. The utility of artificial neural networks and classification and regression trees for the prediction of endometrial cancer in postmenopausal women[J]. Public Health, 2018, 164:1-6.
[10] 翟光宇, 耿炫, 李俊魁, 等. 辅助生殖中胚胎质量和患者年龄对妊娠结局的影响[J]. 中国现代医药杂志, 2017, 19 (11): 34-37.
[11] Takahashi T, Hasegawa A, Igarashi H, et al. Prognostic factors for patients undergoing vitrified-warmed human embryo transfer cycles: a retrospective cohort study[J]. Human Fertility (Camb), 2017, 20(2): 140-146.
[12] von Wolff M, Schwartz AK, Bitterlich N, et al. Only women's age and the duration of infertility are the prognostic factors for the success rate of natural cycle IVF[J]. Archives of Gynecology and Obstetrics, 2019, 299(3): 883-889.
[13] Tannus S, Hatirnaz S, Tan J, et al. Predictive factors for live birth after in vitro maturation of oocytes in women with polycystic ovary syndrome[J]. Archives of Gynecology and Obstetrics, 2018, 297(1): 199-204.
[14] 罗燕群, 刘风华, 易艳红, 等. 年龄、移植胚胎数量、质量与临床妊娠率的关系[J]. 生殖医学杂志, 2014, 23(05): 361-366.
[15] 胡琳莉, 黄国宁, 孙海翔, 等. 辅助生殖技术临床关键指标质控专家共识[J]. 生殖医学杂志, 2018, (09): 828-835.
[16] 任昀, 杨硕, 杨蕊, 等. 促性腺激素释放激素激动剂长方案与拮抗剂方案对体外受精治疗妊娠结局的影响[J]. 北京大学学报(医学版), 2013, 45(6): 877-881.
[17] Dunne C, Lawrence C, Albert A, et al. Longer ovarian stimulation reduces embryo number and clinical pregnancy rate in long GnRH agonist cycles[J]. Minerva Ginecologica, 2017, 69(2): 135-140.
[18] Jiang S, Li L, Li F, et al. Establishment of predictive model for analyzing clinical pregnancy outcome based on IVF-ET and ICSI assisted reproductive technology[J]. Saudi J Biol Sci, 2020, 27(4):1049-1056.
[19] van Loendersloot LL, van Wely M, Repping S, et al. Individualized decision-making in IVF: calculating the chances of pregnancy[J]. Hum Reprod, 2013, 28(11):2972-2980.
[20] 帅健,李丽萍,陈业群.决策树模型与Logistic回归模型在伤害发生影响因素分析中的作用[J].中华疾病控制杂志,2015,19(02):185-189.
[21] Tran D, Cooke S, Illingworth PJ, et al. Deep learning as a predictive tool for fetal heart pregnancy following time-lapse incubation and blastocyst transfer. Hum Reprod. 2019 Jun 4;34(6):1011-1018.
[22] 于医萍, 高一博, 方兰兰, 等. 机器学习在体外受精-胚胎移植技术中的应用[J]. 中华生殖与避孕杂志, 2021, 41(10): 883-892.