目的：探讨结直肠癌患者在接受腹腔镜手术后出现肠梗阻的风险因素，并开发一个预测肠梗阻的 AI 模型。方法：回顾性分析 2020 年 1 月—2023 年 12 月于贵州省人民医院行腹腔镜结直肠癌根治术的 330 例患者的临 床资料，根据术后是否发生肠梗阻为依据，将出现术后肠梗阻的 26 例患者纳入肠梗阻组，未发生肠梗阻的 304 例 患者纳入非肠梗阻组。分别建立随机森林（RF）、梯度提升决策树模型（GBDT）、轻量梯度提升模型（LGBM）、 极端梯度提升模型（XGBoost）、类别型特征梯度提升模型（CatBoost）五种机器学习模型，以受试者工作特征曲线 下面积（AUC）、特异性、敏感性来评价模型的性能，选择最优模型。通过递归特征消除（RFE）筛选最佳的特征 集合，作为腹腔镜手术治疗结直肠癌患者术后并发肠梗阻风险预测因子。结果：研究共纳入 330 例患者，肠梗阻的 发生率为 7.88%。Catboost 模型验证集的 AUC 为 0.862，特异性值为 0.766，整体性能优于其他模型；术前肠梗阻、 手术时间、病灶部位、性别等是影响结直肠癌腹腔镜根治术后合并肠梗阻的高危因素。结论：与目前国内外其他的 机器学习方法相比，CatBoost 可建立更有效的腹腔镜结直肠癌根治术患者术后并发肠梗阻预测模型，具有潜在的应 用价值；术前肠梗阻、手术时间、病灶部位是术后并发肠梗阻的主要影响因素。

Objective: To study the risk factors for postoperative intestinal obstruction in colorectal cancer patients who underwent laparoscopic radical resection, and to develop an AI prediction model for intestinal obstruction. Methods: Clinical data of 330 patients who undenwent laparoscopic radical resection of colorectal cancer in Guizhou Provincian peoples’s Hospital from January 2020 to December 2023 was retrospectively analyzed. They were grouped according to whether intestinal obstruction occurred or not after surgery. A total of 26 patients were enrolled in the intestinal obstruction group, and 304 patients were enrolled in the non-obstruction group. Random forest (RF), gradient boosted decision tree (GBDT), light gradient boosting machine (LGBM), extreme gradient boosting (XGBoost) and categorical features gradient boosting (CatBoost)were established to evaluate the performance of the models using metrics such as area under the curve (AUC), specificity and sensitivity. The best model was selected. Recursive feature elimination (RFE) was used to identify the optimal feature set as predictive factors for postoperative intestinal obstruction in patients who underwent laparoscopic radical resection for colorectal cancer. Results: Among the 330 patients, the incidence rate of intestinal obstruction was 7.88% (26/330). The CatBoost model had an AUC of 0.862 and a specificity of 0.766 in the validation set, showing better overall performance compared to other models. Key risk factors for intestinal obstruction following laparoscopic radical resection of colorectal cancer were previous intestinal obstruction, operative time, treatment method, lesion site and patient gender. Conclusion: Compared to other machine learning methods, the CatBoost method can establish a more effective AI prediction model for postoperative intestinal obstruction in laparoscopic radical resection of colorectal cancer, with potential application value. Preoperative intestinal obstruction, operative time and lesion site were major influencing factors to postoperative intestinal obstruction.

收稿日期：2024-04-18  录用日期：2024-04-29 

Received Date: 2024-04-18  Accepted Date: 2024-04-29 

基金项目：国家自然科学基金面上项目（82172844） 

Foundation Item: National Natural Science Foundation of China (82172844) 

通讯作者：陈莺，Email：chenying126073@126.com 

Corresponding Author: CHEN Ying, Email: chenying126073@126.com 

引用格式：何晓芳，陈洁，李秋萍，等 . 基于机器学习算法的腹腔镜结直肠癌根治术后肠梗阻预测模型 [J]. 机器人外科学杂志（中 英文），2024，5（6）：1205-1210. 

Citation: HE X F, CHEN J, LI Q P, et al. AI prediction model for postoperative intestinal obstruction in patients underwent laparoscopic radical resection of colorectal cancer[J]. Chinese Journal of Robotic Surgery,  2024, 5(6): 1205-1210.

[1] 李政 , 郑晓强 , 王罡 , 等 . 肠梗阻导管对结直肠癌合并肠梗阻患 者术后感染的影响 [J]. 中华医院感染学杂志 , 2021, 31(1): 95-98. 

[2] 许超 , 池畔 . 结直肠癌根治术后肠梗阻的影响因素分析 [J]. 中 华胃肠外科杂志 , 2014, 17(4) : 361-364. 

[3] Quiroga-Centeno A C, Jerez-Torra K A, Martin-Mojica P A, et al. Risk factors for prolonged postoperative ileus in colorectal surgery: a systematic review and meta-analysis[J]. World J Surg, 2020, 44(5): 1612-1626. 

[4] 杨玉兵 , 邢文英 , 王耿泽 . 结直肠癌根治术后发生肠梗阻的 影响因素分析 1686 例 [J]. 世界华人消化杂志 , 2015, 23(10): 1664-1669. 

[5] 白雪杉 , 林国乐 . 2019. V1 版《NCCN 结直肠癌诊治指南》更 新要点解析 [J]. 中国全科医学 . 2019, 22(33): 4031-4034. 

[6] 陈孝平 , 汪建平 , 赵继宗 . 外科学 [M]. 人民卫生出版社 , 2013. 

[7] GONG H W, WANG M Y, ZHANG H X, et al. An explainable AI approach for the rapid diagnosis of COVID-19 using ensemble learning algorithms[J]. Front Public Health, 2022. DOI: 10.3389/ fpubh.2022.874455. 

[8] Prokhorenkova L, Gusev G, Vorobev A, et al. CatBoost: unbiased boosting with categorical features[J]. Advances in neural information processing systems, 2018. DOI: org/10.48550/arXiv.1706.09516. 

[9] CHEN Q, MENG Z P, LIU X Y, et al. Decision variants for the automatic determination of optimal feature subset in RF-RFE[J]. Genes(Basel), 2018, 9(6): 301.

[10] GONG H W, YOU X, JIN M, et al. Graph neural network and multidata heterogeneous networks for microbe-disease prediction[J]. Front Microbiol, 2022. DOI: 10.3389/fmicb.2022.1077111. 

[11] Zeinali F, Stulberg J J, Delaney C P. Pharmacological management of postoperative ileus[J]. Can J Surg, 2009, 52(2): 153-157. 

[12] Bragg D, El-Sharkawy A M, Psaltis E, et al. Postoperative ileus: Recent developments in pathophysiology and management[J]. Clin Nutr, 2015, 34(3): 367-376. 

[13] Kronberg U, Kiran R P, Soliman M S M, et al. A characterization of factors determining postoperative ileus after laparoscopic colectomy enables the generation of a novel predictive score[J]. Ann Surg, 2011, 253(1): 78-81. 

[14] Moghadamyeghaneh Z, Hwang G S, Hanna M H, et al. Risk factors for prolonged ileus following colon surgery[J]. Surg Endosc, 2016, 30(2): 603-609. 

[15] Goldstone R N, Popowich D A. Laparoscopic intracorporeal anastomosis[J]. Clin Colon Rectal Surg, 2023, 36(1): 74-82. 

[16] Honjo K, Kawai M, Tsuchiya Y, et al. Risk factors for small-bowel obstruction after colectomy for colorectal cancer: a retrospective study[J]. Surg Today, 2023, 53(9): 1038-1046. 

[17] Kuruba R, Fayard N, Snyder D. Epidural analgesia and laparoscopic technique do not reduce incidence of prolonged ileus in elective colon resections[J]. Am J Surg, 2012, 204(5): 613-618. 

[18] Wolthuis A M, Bislenghi G, Fieuws S, et al. Incidence of prolonged postoperative ileus after colorectal surgery: a systematic review and meta-analysis[J]. Colorectal Dis, 2016, 18(1): 1-9. 

[19] ZHENG H D, LIU Y R, CHEN Z Z, et al. Novel nomogram for predicting risk of early postoperative small bowel obstruction after right colectomy for cancer[J]. World J Surg Oncol, 2022, 20(1): 19. 

[20] WANG X J, CHI P, LIN H M, et al. Risk factors for early postoperative small bowel obstruction after elective colon cancer surgery: an observational study of 1, 244 consecutive patients[J]. Dig Surg, 2018, 35(1): 49-54. 

[21] DAI X J, GE X L, YANG J B, et al. Increased incidence of prolonged ileus after colectomy for inflammatory bowel diseases under ERAS protocol: a cohort analysis[J]. J Surg Res, 2017. DOI: 10.1016/ j.jss.2016.12.031. 

[22] Yanagisawa T, Tatematsu N, Horiuchi M, et al. Prolonged preoperative sedentary time is a risk factor for postoperative ileus in patients with colorectal cancer: a propensity score-matched retrospective study[J]. Support Care Cancer, 2023, 32(1): 54. 

[23] JU H, SHEN K, LI J X, et al. Total postoperative opioid dose is an independent risk factor for prolonged postoperative ileus after laparoscopic colorectal surgery: a case-control study[J]. Korean J Anesthesiol, 2024, 77(1): 133-138. 

[24] Seo S H B, Carson D A, Bhat S, et al. Prolonged postoperative ileus following right-versus left-sided colectomy: A systematic review and meta-analysis[J]. Colorectal Dis, 2021, 23(12): 3113-3122. 

[25] Masoomi H, Kang C Y, Chaudhry O, et al. Predictive factors of early bowel obstruction in colon and rectal surgery: data from the Nationwide Inpatient Sample, 2006-2008[J]. J Am Coll Surg, 2012, 214(5): 831-837. 

[26] Chapuis P H, Bokey L, Keshava A, et al. Risk factors for prolonged ileus after resection of colorectal cancer: an observational study of 2400 consecutive patients[J]. Ann Surg, 2013, 257(5): 909-915. 

[27] Fujiyoshi S, Homma S, Yoshida T, et al. A Study of risk factors of postoperative ileus after laparoscopic colorectal resection[J]. Ann Gastroenterol Surg, 2023, 7(6): 949-954. 

[28] Lambrichts D P V, Boersema G S A, Tas B, et al. Nicotine chewing gum for the prevention of postoperative ileus after colorectal surgery: a multicenter, double-blind, randomised, controlled pilot study[J]. Int J Colorectal Dis, 2017, 32(9): 1267-1275.