[1]王 晶,郝晓燕,白丛霞,等.机器学习算法挖掘整合临床实验室多参数检验大数据构建结直肠息肉风险预测模型及效能评估[J].现代检验医学杂志,2026,41(03):98-103.[doi:10.3969/j.issn.1671-7414.2026.03.018]
　WANG Jing,HAO Xiaoyan,BAI Congxia,et al.Machine Learning Algorithm-Based Mining and Integration of Clinical Laboratory Multi-Parameter Big Data for the Construction and Performance Evaluation of a Colorectal Polyps Risk Prediction Model[J].Journal of Modern Laboratory Medicine,2026,41(03):98-103.[doi:10.3969/j.issn.1671-7414.2026.03.018]
点击复制

机器学习算法挖掘整合临床实验室多参数检验大数据构建结直肠息肉风险预测模型及效能评估()

《现代检验医学杂志》[ISSN:/CN:]

卷:
第41卷
期数:
2026年03期
页码:
98-103
栏目:
论著
出版日期:
2026-05-13

文章信息/Info

Title:
Machine Learning Algorithm-Based Mining and Integration of Clinical Laboratory Multi-Parameter Big Data for the Construction and Performance Evaluation of a Colorectal Polyps Risk Prediction Model
文章编号:
1671-7414（2026）03-098-06
作者:
王 晶a，郝晓燕a，白丛霞a，邓 娟b，周 磊a，王 娟a，刘家云a
空军军医大学西京医院a.检验科；b.消化内科，西安 710032
Author(s):
WANG Jinga，HAO Xiaoyana，BAI Congxiaa，DENG Juanb，ZHOU Leia，WANG Juana，LIU Jiayuna
a. Department of Labo-ratory Medicine; b. Department of Gastroenterology, Xijing Hospital of Air Force Medical University, Xi’an 710032, China
关键词:
结直肠息肉; 机器学习算法; 风险预测模型
分类号:
R574；R446.1
DOI:
10.3969/j.issn.1671-7414.2026.03.018
文献标志码:
A
摘要:
目的?通过机器学习算法挖掘常规检验大数据构建结直肠息肉预测模型，为结直肠癌癌前病变的早期诊断和预防性治疗提供新思路。方法本研究收集空军军医大学西京医院2023年10月~2024年12月诊断为结直肠息肉的患者423例，以及健康受试者305例。收集年龄、性别及实验室检测指标共66项，采用拉索交叉验证（LassoCV）、极端梯度提升递归特征消减交叉验证（XGBoost-RFECV）、支持向量机递归特征消除交叉验证（SVM-RFECV）三种方法进行特征筛选。应用合成少数类过采样技术（SMOTE）对数据进行平衡处理，处理后共846例，使用随机森林（RandomForest）、高斯朴素贝叶斯（GNB）、决策树（DecisionTree）、逻辑回归（LogisticRegression）、梯度提升决策树（GBDT）、补集朴素贝叶斯（CNB）、自适应提升（AdaBoost）、极端梯度提升（XGBoost）和轻量级梯度提升（LightGBM）9种机器学习算法构建结直肠息肉早期预测模型，其中760例为训练集，86例为验证集用于验证模型的预测效能。结果经筛选最终纳入活化部分凝血活酶时间（APTT）、凝血酶时间（TT）、凝血酶原时间（PT）、平均血小板体积（MPV）、血小板分布宽度（PDW）、直接胆红素（DBIL）、尿素（BU）、钠（Na+）共8个特征（均P<0.05）用于构建结直肠息肉预测模型。采用9种机器学习方法构建的模型中，XGBoost算法模型预测效果最佳，诊断曲线下面积[AUC（95%CI）]为0.961（0.951～0.971），灵敏度0.894，特异度为0.889。结论XGBoost算法联合8种检验指标建立的结直肠息肉风险预测模型，为肠镜检查的推广、结直肠癌的早筛早诊提供了新依据。
Abstract:
Objective To establish a prediction model for colorectal polyps via mining big data derived from routine laboratory ex-aminations through machine learning algorithms, and to provide novel insights for the early diagnosis and prophylactic treatment of precancerous lesions of colorectal cancer. Method A total of 423 patients diagnosed with colorectal polyps and 305 healthy controls were enrolled at Xijing Hospital of Air Force Medical University between October 2023 and December 2024. Sixty-six variables, including age, gender, and laboratory test indicators were collected. Feature selection was performed using three dis-tinct methods: least absolute shrinkage and selection operator cross-validation (LassoCV), extreme gradient boosting-recursive feature elimination cross-validation (XGBoost-RFECV), and support vector machine-recursive feature elimination cross-valida-tion (SVM-RFECV). The Synthetic Minority Oversampling Technique (SMOTE) was applied to balance the dataset, resulting in a total of 846 eligible cases after processing. Subsequently, nine machine learning algorithms, namely Random Forest, Gaussian Naive Bayes (GNB), Decision Tree, Logistic Regression, gradient boosting decision tree (GBDT), complement naive Bayes (CNB), Adaptive Boosting (AdaBoost), extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM), were utilized to establish an early prediction model for colorectal polyps. Among all cases, 760 were assigned to the training set and 86 to the validation set for evaluating the predictive performance of the model. Results A total of eight variables was finally incorporated into the colorectal polyp prediction model, including activated partial thromboplastin time (APTT), thrombin time (TT), prothrombin time (PT), mean platelet volume (MPV), platelet distribution width (PDW), direct bilirubin (DBIL), urea (BU) and sodium ion (Na+). Among the nine machine learning algorithms evaluated, the XGBoost model exhibited the optimal predictive efficacy, with an area under the receiver operating characteristic curve, with an AUC(95%CI) of 0.961 (0.951～0.971), a sensi-tivity of 0.894, and a specificity of 0.889. Conclusions The risk prediction model for colorectal polyps, developed by integrating the XGBoost algorithm with eight laboratory indicators, provides a new evidence base for the popularization of colonoscopy and the early screening and diagnosis of colorectal cancer.

参考文献/References:

[1] SUNG H, FERLAY J, SIEGEL R L, et al. Global cancer statistics 2020: GLOBOCAN estimates of inci-dence and mortality worldwide for 36 cancers in 185 countries[J]. CA-A Cancer Journal for Clinicians, 2021, 71(3): 209-249.
[2] 郑荣寿,陈茹,韩冰峰,等.2022年中国恶性肿瘤流行情况分析[J].中华肿瘤杂志, 2024, 46(3): 221-231. ZHENG R S, CHEN R, HAN B F, et al. Cancer inci-dence and mortality in China, 2022[J]. Chinese Journal of Oncology, 2024, 46(3): 221-231.
[3] ZAUBER A G, WINAWER S J, O’BRIEN M J, et al. Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths[J]. New England Journal of Medicine, 2012, 366(8): 687-696.
[4] 杨婷,董泽华,李佳,等.结直肠黏膜癌前状态及癌前病变内镜下随访策略[J].中华消化杂志, 2024, 44(7):483-486. YANG T, DONG Z H, LI J, et al. Endoscopic follow-up strategies for colorectal precancerous conditions and precancerous lesions[J]. Chinese Journal of Digestion, 2024, 44(7): 483-486.
[5] JIANG T, GRADUS J L, LASH T L, et al. Addressing measurement error in random forests using quantitative bias analysis[J]. American Journal of Epidemiology, 2021, 190(9): 1830-1840.
[6] EMMANUEL T, MAUPONG T, MPOELENG D, et al. A survey on missing data in machine learning [J]. Jour-nal of Big Data, 2021, 8(1): 140.
[7] 郝晓燕,周磊,白丛霞,等.基于机器学习方法建立血清多种肿瘤标志物联合区分胃炎和胃癌的三种模型及评价[J].现代检验医学杂志, 2023, 38(2): 124-128. HAO X Y, ZHOU L, BAI C X, et al. Establishment and evaluation of three models for the distinguish gastritis and gastric cancer with multiple serum tumor markers based on machine learning methods[J]. Journal of Mod-ern Laboratory Medicine, 2023, 38(2): 124-128.
[8] BARTLETT P L, LONG P M, LUGOSI G, et al. Be-nign overfitting in linear regression[J]. Proceedings of the National Academy of Sciences of the United States of American, 2020, 117(48): 30063-30070.
[9] 李丹,汪嵘.结直肠癌肠镜筛查参检率影响因素研究[J].肿瘤研究与临床, 2024, 36(7): 557-560. LI D, WANG R. Factors influencing the participation rate in colorectal cancer screening by colonoscopy[J]. Cancer Research and Clinic, 2024, 36(7): 557-560.
[10] 李怡玲,关旭,窦利州,等.粪便多靶点DNA与粪便潜血联合检测技术对结直肠癌早诊早筛的诊断价值分析[J].中华医学杂志, 2022, 102(33): 2607-2613. LI Y L, GUAN X, DOU L Z, et al. The clinical value of multi-target stool fecal immunochemical test-DNA in early screening and diagnosis for colorectal cancer[J]. National Medical Journal of China, 2022, 102(33): 2607-2613.
[11] 朱小伶,贺雪,刘雪莲,等.品管圈在降低体检中心大便检查弃检率中的效果[J].当代护士, 2018, 25(32):176-177. ZHU X L, HE X, LIU X L, et al. The effect of Quality Control Circle in reducing the abandonment rate of stool tests in the physical examination center[J]. Mod-ern Nurse, 2018, 25(32): 176-177.
[12] ZUO D, YANG L X, JIN Y, et al. Machine learn-ing-based models for the prediction of breast cancer recurrence risk[J]. BMC Medical Informatics and Deci-sion Making, 2023, 23(1): 276.
[13] XIONG Y B, MA Y, RUAN L G, et al. Comparing different machine learning techniques for predicting COVID-19 severity[J]. Infectious Diseases of Poverty, 2022, 11(1): 19.
[14] ZAREI GHOBADI M, AFSANEH E, EMAMZADEH R. Gene biomarkers and classifiers for various subtypes of HTLV-1-caused ATLL cancer identified by a com-bination of differential gene co-expression and support vector machine algorithms[J]. Medical Microbiology and Immunology, 2023, 212(4): 263-270.
[15] LIANG D, WANG L, ZHONG P P, et al. Perspective:global burden of iodine deficiency: insights and projec-tions to 2050 using XGBoost and SHAP [J]. Advances in Nutrition, 2025, 16(3): 100384.
[16] ZHANG L, YE J, LUO Q Y, et al. Prediction of poor outcomes in patients with colorectal cancer: elevated preoperative prothrombin time (PT) and activated par-tial thromboplastin time (APTT) [J]. Cancer Manage-ment and Research, 2020, 12: 5373-5384.
[17] LEBLANC R, PEYRUCHAUD O. Metastasis: new functional implications of platelets and megakaryo-cytes[J]. Blood, 2016, 128(1): 24-31.
[18] CHEN Z J, VONG C T, GAO C F, et al. Bilirubin nano-medicines for the treatment of reactive Oxygen species (ROS)-Mediated diseases[J]. Molecular Pharmaceutics, 2020, 17(7): 2260-2274.
[19] WU E, WEI G F, LI Y, et al. Serum urea concentration and risk of 16 site-specific cancers, overall cancer, and cancer mortality in individuals with metabolic syn-drome: a cohort study[J]. BMC Medicine, 2024, 22(1):536.
[20] POKU L O, PHIL M, CHENG Y N, et al. 23Na-MRI as a noninvasive biomarker for cancer diagnosis and prognosis [J]. Journal of Magnetic Resonance Imaging, 2021, 53(4): 995-1014.
[21] 张颖.结直肠癌患者血清SDC4,CXCL1水平检测在临床诊断及淋巴结转移评估中的应用价值[J].现代检验医学杂志, 2025, 40(1): 116-121. ZHANG Y. Application value of serum SDC4,CXCL1 level detection in clinical diagnosis and lymph node me-tastasis assessment in colorectal cancer patients[J]. Journal of Modern Laboratory Medicine, 2025, 40(1): 116-121.
[22] 付超英,黄士月,张云霞.直肠癌组织中LncRNA TTN-AS1及角鲨烯环氧酶表达与临床病理特征及预后相关性研究[J].现代检验医学杂志, 2024, 39(5):12-16, 53. FU C Y, HUANG S Y, ZHANG Y X. Correlation between the expression of LncRNA TTN-AS1 and squalene epoxidase in rectal cancer tissue and clinico-pathological characteristics and prognosis[J]. Journal of Modern Laboratory Medicine, 2024, 39(5): 12-16, 53.

相似文献/References:

[1]张博超,李 强,张 如,等.基于GEO数据库应用WGCNA和机器学习算法筛选肝细胞癌相关核心基因及验证[J].现代检验医学杂志,2026,41(01):86.[doi:10.3969/j.issn.1671-7414.2026.01.017]
　ZHANG Bochao,LI Qiang,ZHANG Ru,et al.Screening of Core Genes Related to Hepatocellular Carcinoma Using WGCNA and Machine Learning Algorithm Based on GEO Database and Verification[J].Journal of Modern Laboratory Medicine,2026,41(03):86.[doi:10.3969/j.issn.1671-7414.2026.01.017]

备注/Memo

备注/Memo:
基金项目：国家重点研发项目（编号：2022YFC3602302）。
作者简介：王晶（1993-），女，硕士，主管技师，主要从事临床检验与分子生物学，E-mail：wjmed0429@163.com。
通讯作者：刘家云（1971-），男，教授，主任医师，博士研究生导师，E-mail：jiayun@fmmu.edu.cn。
更新日期/Last Update: 2026-05-15