基于机器学习建立术前预测近端胃癌食管切缘阳性模型

doi:10.6040/j.issn.1671-7554.0.2024.0292

摘要/Abstract

摘要： 目的建立术前预测近端胃癌食管切缘阳性的机器学习模型,并比较其与传统Logistics模型的预测性能。方法回顾性分析2013年1月至2022年12月于衡水市人民医院胃肠外科接受近端胃癌手术的382例患者的临床病理资料,根据食管切缘状态分为切缘阳性组(n=30)和切缘阴性组(n=352)。将研究对象按2∶1比例随机分为训练集(n=254)和测试集(n=128),采用合成少数样本过采样技术(synthetic minority oversampling technique, SMOTE)处理训练集中的不平衡数据,基于平衡后SMOTE数据集建立随机森林(random forest, RF)、支持向量机(support vector machine, SVM)和极端梯度提升(extreme gradient boosting, Xgboost)3种机器学习模型及Logistic回归模型。通过上述4种模型,在测试集中预测食管切缘阳性时,利用受试者操作特征曲线下面积(area under curve, AUC)数值来比较不同模型的预测性能,对最佳预测模型中预测因素的重要性进行可视化排序。结果 4种模型的AUC值从高到低依次为RF模型0.772(95%CI:0.620~0.925),SVM模型0.747(95%CI:0.604~0.891),Logistic回归模型0.716(95%CI:0.537~0.895)和Xgboost模型0.710(95%CI:0.560~0.859)。RF模型预测性能最佳。肿瘤大小、肿瘤位置、Borrmann分型、Lauren分型及cT分期是RF模型中前5位重要因素。结论所建立的术前预测近端胃癌食管切缘阳性的RF模型性能良好;肿瘤大小、肿瘤位置、Borrmann分型、Lauren分型及cT分期是主要的预测因素。

关键词: 胃癌, 进展期, 切缘阳性, 机器学习, 预测模型

Abstract: Objective To develop machine learning models for preoperative prediction of positive esophageal margins in proximal gastric cancer and to compare its prediction performance with conventional Logistics models. Methods A total of 382 patients with proximal gastric cancer who received operation at the Department of Gastrointestinal Surgery of Hengshui Peoples Hospital from January 2013 to December 2022 were retrospectively analyzed and divided into the margin-positive group(n=30)and the margin-negative group(n=352)according to the pathologic diagnosis. The clinicopathological factors that might affect the positive esophageal margins of proximal gastric cancer were collected, and the study population were randomly divided into the training set(n=254)and the test set(n=128)in a ratio of 2∶1. The unbalanced data in the training set were processed by synthetic minority oversampling technique(SMOTE). Three machine learning models, that is, the random forest(RF), support vector machine(SVM)and extreme gradient boosting(Xgboost), and Logistic regression model were established based on the balanced SMOTE dataset. The predictive performance of the different models was compared by the AUC values of the above four models in predicting positive esophageal margins in the test set, and the importance of the predictors in the best predictive model was visually ranked. Results RF had the highest AUC value(0.772, 95%CI: 0.620-0.925), followed by SVM(AUC: 0.747, 95%CI: 0.604-0.891), the Logistic regression(AUC: 0.716, 95%CI: 0.537-0.895), and Xgboost(AUC: 0.710, 95%CI: 0.560-0.859). The RF model had the best predictive performance. Tumor size, tumor location, Borrmann staging, Lauren staging and cT staging were the top 5 important factors in the RF model. Conclusion The established random forest model for preoperative prediction of positive margins in proximal gastric cancer shows good performance, with tumor size, tumor location, Borrmann staging, Lauren staging and cT staging being the main predictive factors.

Key words: Gastric cancer, Advanced, Positive proximal margin, Machine learning, Predictive model

中图分类号:

R735.2

郭振江,王宁,赵光远,杜立强,崔朝勃,刘防震. 基于机器学习建立术前预测近端胃癌食管切缘阳性模型[J]. 山东大学学报 (医学版), 2024, 62(7): 78-83.

GUO Zhenjiang, WANG Ning, ZHAO Guangyuan, DU Liqiang, CUI Zhaobo, LIU Fangzhen. Development of preoperative models for predicting positive esophageal margin in proximal gastric cancer based on machine learning[J]. Journal of Shandong University (Health Sciences), 2024, 62(7): 78-83.

参考文献

[1] 汪欣, 陈晓. 近端胃癌的诊断与治疗进展[J]. 中华普通外科杂志, 2023, 38(4): 241-244. WANG Xin, CHEN Xiao. Progress in diagnosis and treatment of proximal gastric cancer[J]. Chinese Journal of General Surgery, 2023, 38(4): 241-244.
[2] Han WH, Eom BW, Yoon HM, et al. The optimal extent of lymph node dissection in gastroesophageal junctional cancer: retrospective case control study[J]. BMC Cancer, 2019, 19(1): 719. doi:10.1186/s12885-019-5922-8.
[3] Cho BC, Jeung HC, Choi HJ, et al. Prognostic impact of resection margin involvement after extended(D2/D3)gastrectomy for advanced gastric cancer: a 15-year experience at a single institute[J]. J Surg Oncol, 2007, 95(6): 461-468.
[4] Talavera-Urquijo E, Davies AR, Wijnhoven BPL. Prevention and treatment of a positive proximal margin after gastrectomy for cardia cancer[J]. Updates Surg, 2023, 75(2): 335-341.
[5] Jiang ZY, Liu CY, Cai ZL, et al. Impact of surgical margin status on survival in gastric cancer: a systematic review and meta-analysis[J]. Cancer Control, 2021, 28: 10732748211043665. doi:10.1177/10732748211043665.
[6] Nakanishi K, Morita S, Taniguchi H, et al. Diagnostic accuracy and usefulness of intraoperative margin assessment by frozen section in gastric cancer[J]. Ann Surg Oncol, 2019, 26(6): 1787-1794.
[7] Watanabe A, Adamson H, Lim H, et al. Intraoperative frozen section analysis of margin status as a quality indicator in gastric cancer surgery[J]. J Surg Oncol, 2023, 127(1): 66-72.
[8] Berlth F, Kim WH, Choi JH, et al. Prognostic impact of frozen section investigation and extent of proximal safety margin in gastric cancer resection[J]. Ann Surg, 2020, 272(5): 871-878.
[9] Bissolati M, Desio M, Rosa F, et al. Risk factor analysis for involvement of resection margins in gastric and esophagogastric junction cancer: an Italian multicenter study[J]. Gastric Cancer, 2017, 20(1): 70-82.
[10] De Manzoni G, Marrelli D, Baiocchi GL, et al. The Italian Research Group for Gastric Cancer(GIRCG)guidelines for gastric cancer staging and treatment: 2015[J]. Gastric Cancer, 2017, 20(1): 20-30.
[11] Kumazu Y, Hayashi T, Yoshikawa T, et al. Risk factors analysis and stratification for microscopically positive resection margin in gastric cancer patients[J]. BMC Surg, 2020, 20(1): 95. doi:10.1186/s12893-020-00744-5.
[12] Popovic D, Glisic T, Milosavljevic T, et al. The importance of artificial intelligence in upper gastrointestinal endoscopy[J]. Diagnostics, 2023, 13(18): 2862. doi:10.3390/diagnostics13182862.
[13] 巨艳丽, 王丽华, 成芳, 等. 基于机器学习构建放射性碘治疗疗效的预测模型[J]. 山东大学学报(医学版), 2023, 61(1): 94-99. JU Yanli, WANG Lihua, CHENG Fang, et al. Construction of predictive models of radioiodine therapy based on machine learning[J]. Journal of Shandong University(Health Sciences), 2023, 61(1): 94-99.
[14] 何文琪, 伍兵. 胃癌术前cTN分期的影像研究进展[J]. 国际医学放射学杂志, 2019, 42(1): 76-80. HE Wenqi, WU Bing. Progress in imaging studies on preoperative cTN staging of gastric cancer[J]. International Journal of Medical Radiology, 2019, 42(1): 76-80.
[15] Hassan A, Gulzar Ahmad S, Ullah Munir E, et al. Predictive modelling and identification of key risk factors for stroke using machine learning[J]. Sci Rep, 2024, 14(1): 11498. doi:10.1038/s41598-024-61665-4.
[16] Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent[J]. J Stat Softw, 2010, 33(1): 1-22.
[17] 刘颂, 夏秋媛, 付尧, 等. 胃癌根治术中食管切缘跳跃转移30例临床特征分析[J]. 中华胃肠外科杂志, 2023, 26(7): 675-679. LIU Song, XIA Qiuyuan, FU Yao, et al. Skip metastasis at the esophageal resection margin in radical gastrectomy: clinical characteristics of 30 cases[J]. Chinese Journal of Gastrointestinal Surgery, 2023, 26(7): 675-679.
[18] Haverkamp L, Ruurda JP, van Leeuwen MS, et al. Systematic review of the surgical strategies of adenocarcinomas of the gastroesophageal junction[J]. Surg Oncol, 2014, 23(4): 222-228.
[19] Fuchs H, Hölscher AH, Leers J, et al. Long-term quality of life after surgery for adenocarcinoma of the esophagogastric junction: extended gastrectomy or transthoracic esophagectomy?[J]. Gastric Cancer, 2016, 19(1): 312-317.
[20] Brown AM, Giugliano DN, Berger AC, et al. Surgical approaches to adenocarcinoma of the gastroesophageal junction: the Siewert II conundrum[J]. Langenbecks Arch Surg, 2017, 402(8): 1153-1158.
[21] Imamura Y, Watanabe M, Oki E, et al. Esophagogastric junction adenocarcinoma shares characteristics with gastric adenocarcinoma: literature review and retrospective multicenter cohort study[J]. Ann Gastroenterol Surg, 2021, 5(1): 46-59.
[22] Juez LD, Barranquero AG, Priego P, et al. Influence of positive margins on tumour recurrence and overall survival after gastrectomy for gastric cancer[J]. ANZ J Surg, 2021, 91(7/8): E465-E473.
[23] Tu RH, Lin JX, Wang W, et al. Pathological features and survival analysis of gastric cancer patients with positive surgical margins: a large multicenter cohort study[J]. Eur J Surg Oncol, 2019, 45(12): 2457-2464.
[24] Pang T, Nie MM, Yin K. The correlation between the margin of resection and prognosis in esophagogastric junction adenocarcinoma[J]. World J Surg Oncol, 2023, 21(1): 316. doi:10.1186/s12957-023-03202-7.
[25] Jung MK, Schmidt T, Chon SH, et al. Current surgical treatment standards for esophageal and esophagogastric junction cancer[J]. Ann N Y Acad Sci, 2020, 1482(1): 77-84.
[26] van der Werf LR, Cords C, Arntz I, et al. Population-based study on risk factors for tumor-positive resection margins in patients with gastric cancer[J]. Ann Surg Oncol, 2019, 26(7): 2222-2233.
[27] Fukagawa T, Katai H, Mizusawa J, et al. A prospective multi-institutional validity study to evaluate the accuracy of clinical diagnosis of pathological stage III gastric cancer(JCOG1302A)[J]. Gastric Cancer, 2018, 21(1): 68-73.
[28] Nechita Vi, Al-Hajjar N, Leucuta DC, et al. Inflammatory ratios as survival prognostic factors in resectable gastric adenocarcinoma[J]. Diagnostics(Basel), 2023, 13(11): 1910. doi:10.3390/diagnostics13111910.

相关文章 15

[1]	廖原,门丹,李一帆,李怀臣,龙飞,刘铱. PM_2.5短期暴露对结核病发病风险的个体精准评估[J]. 山东大学学报 (医学版), 2026, 64(3): 116-123.
[2]	刘禹,霍娅娅,龚丞,梁婷,李斌. 基于机器学习的骨科生物材料设计与优化[J]. 山东大学学报 (医学版), 2026, 64(2): 22-33.
[3]	王建民,李晓峰,由志涛,董圣杰,赵宇驰,李占菊,邹德鑫,张剑锋,孙涛,杜伟. 基于可解释机器学习的后路腰椎椎体间融合术后慢性疼痛风险预测模型构建[J]. 山东大学学报 (医学版), 2026, 64(2): 78-88.
[4]	孙爽爽,仉率杰,张伯韬,袁莹,于媛媛,薛付忠. 基于真实世界研究的18~50岁人群急性缺血性卒中影响因素[J]. 山东大学学报 (医学版), 2025, 63(9): 40-46.
[5]	申路佳,逯天威,巩伟明,赵岩松,王淑康,袁中尚. 代谢风险评分在2型糖尿病人群心血管结局预测中的应用[J]. 山东大学学报 (医学版), 2025, 63(8): 69-78.
[6]	王丽云,高天勤,刘雨佳,陈青,陈柳,沙凯辉. 基于机器学习产后压力性尿失禁风险预测模型的构建及验证[J]. 山东大学学报 (医学版), 2025, 63(6): 55-66.
[7]	陈瑛翼,游倩,王意,张帆,李凤,季舒铭,徐浩源,饶志勇. 办公室职员肌肉质量减少预测模型的开发与验证[J]. 山东大学学报 (医学版), 2025, 63(4): 26-35.
[8]	王宝炫,焦杰,张厚君,刘奇,于冠英. 衰弱与肌少症评估在胃肠道肿瘤术后结局预测中的应用与展望[J]. 山东大学学报 (医学版), 2025, 63(4): 51-58.
[9]	张洁,赵颖慧,董雅琪,李娟,李培龙,杜鲁涛. GPR133基因甲基化在早期胃癌淋巴结转移预测中的作用[J]. 山东大学学报 (医学版), 2025, 63(3): 76-84.
[10]	崔倩倩,李金鹏,吴豫丹,侯志平,孙玮螺,崔盼盼,何培元. 粪便miRNA对结直肠进展期腺瘤无创筛查的临床价值[J]. 山东大学学报 (医学版), 2025, 63(2): 43-50.
[11]	陈文亮,王欢欢,郝金锦,弓蕊,赵强,张飞,高磊,董静逊. lncRNA PVT1表达在胃癌预后评估及恶性进展中的作用:基于列线图模型与细胞功能实验的研究[J]. 山东大学学报 (医学版), 2025, 63(10): 61-71.
[12]	李永,崔书君,杨飞,张凡,殷晓霞. 基于增强MRI的亚区域影像组学模型可预测乳腺癌患者新辅助化疗后的病理完全反应[J]. 山东大学学报 (医学版), 2025, 63(1): 81-89.
[13]	孙丽娜,白红艳,牛宗格,张福帅,曲仪庆. 基于SII构建及评价预测ARDS住院死亡率的在线临床风险模型[J]. 山东大学学报 (医学版), 2024, 62(7): 10-20.
[14]	王静,刘晓菲,曾荣,许长娟,张锦涛,董亮. 基于机器学习算法鉴定哮喘的坏死性凋亡相关生物标志物[J]. 山东大学学报 (医学版), 2024, 62(7): 21-32.
[15]	郭鑫,孟君,郑世良,董秀红. 老年胃癌患者衰弱与人体成分的相关性[J]. 山东大学学报 (医学版), 2024, 62(4): 40-47.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed