基于复合核偏最小二乘模型的胶质瘤多模态 MRI 数据预测 MGMT 启动子甲基化合并 IDH 突变*

杨琪 ,  闫俊羽 ,  刘夏 ,  王彤 ,  房瑞玲 ,  曾平 ,  杨海涛 ,  罗艳虹 ,  余红梅 ,  张岩波 ,  杨国强 ,  曹红艳

中国卫生统计 ›› 2026, Vol. 43 ›› Issue (4) : 482 -488.

PDF (1564KB)
中国卫生统计 ›› 2026, Vol. 43 ›› Issue (4) : 482 -488. DOI: 10.11783/j.issn.1002-3674.2026.04.001
论著

基于复合核偏最小二乘模型的胶质瘤多模态 MRI 数据预测 MGMT 启动子甲基化合并 IDH 突变*

作者信息 +

The Prediction of MGMT Promoter Methylation Combined with IDH Mutation in Glioma Using a Composite Kernel Partial Least Squares Model based on Multimodal MRI Data

Author information +
文章历史 +
PDF (1601K)

摘要

目的 探讨基于术前MRI影像特征,构建胶质瘤中MGMT启动子甲基化合并IDH 突变的预测模型,为胶质瘤患者的预后评估与治疗策略提供理论依据。方法 收集来自TCGA/TCIA 公共数据库的208 例胶质瘤患者的术前 MRI 图像、临床及基因信息。根据MGMT 启动子甲基化和IDH 突变状态,将患者分为IDHmut&MGMTmet 亚型与Other 亚型,并比较两亚型间各因素的差异,分析分子亚型与生存结局的关系。从术前MRI 图像中提取并量化CE-T1 和T2-FLAIR 序列的影像学特征,经标准化后,采用惩罚logistic 回归筛选特征,使用基于遗传算法的融合核偏最小二乘算法 (fused kernel partial least squares with the genetic algorithm, GA-fKPLS)、K 近邻(K-nearest neighbor, KNN)、最小绝对收缩和选择算法(least absolute shrinkage and selection operator, LASSO)、逻辑回归(logistic regression, LR)和朴素贝叶斯 (Naïve Bayes, NB)构建胶质瘤亚型预测模型,通过AUC、灵敏度、特异度、准确率、约登指数、G 均值、马修斯相关系数和F值评估模型性能。结果 胶质瘤不同分子亚型在年龄、肿瘤病理分级、生存状况、MGMT启动子甲基化状态以及IDH突变 状态均存在显著差异。Other 亚型患者的死亡风险是IDHmut&MGMTmet 亚型的2.675 倍。通过特征筛选得到42 个CE-T1 特征和68 个T2-FLAIR 特征用于构建预测模型。GA-fKPLS 预测模型在多项评价指标上均高于其他四种模型。结论 基于术前多模态MRI数据构建的GA-fKPLS模型在预测胶质瘤患者分子亚型方面表现最佳,可为胶质瘤患者的治疗决策、预后评估提供重要临床辅助价值。

Abstract

Objective To explore the construction of a predictive model for the identification of MGMT promoter methylation combined with IDH mutation in glioma based on preoperative MRI imaging features, providing theoretical reference for prognosis assessment and treatment strategy of glioma patients.Methods Preoperative MRI images, clinical data, and genetic data of 208 glioma patients were collected from the TCGA/TCIA public databases. Based on IDH mutation and MGMT promoter methylation status, patients were classified into the IDHmut&MGMTmet subtype and the Other subtype. Differences in various factors between the two subtypes were compared, and the relationship between molecular subtypes and survival outcomes was analyzed. Radiomic features of CE-T1 and T2-FLAIR sequences were extracted and quantified from the preoperative MRI images. After feature standardization, penalized logistic regression was used for feature selection. Predictive models for glioma subtypes were built using fused kernel partial least squares with the genetic algorithm (GA-fKPLS), K-nearest neighbor (KNN), least absolute shrinkage and selection operator (LASSO), logistic regression (LR), and Naïve Bayes (NB). Model performance was evaluated based on AUC, sensitivity, specificity, accuracy, Youden’s index, G-mean, Matthews correlation coefficient, and F-measure.Results Significant differences in age, tumor pathological grade, survival status, MGMT promoter methylation, and IDH mutation status were found between the two molecular glioma subtypes. The death risk of the Other subtype was 2. 675 times higher than glioma patients of the IDHmut&MGMTmet subtype. 42 features from CE-T1 sequence and 68 features from T2-FLAIR sequence were selected for model construction. The GA-fKPLS model outperforms the other four models across multiple evaluation metrics. Conclusion The GA-fKPLS model based on preoperative MRI features exhibits the best performance in predicting the molecular subtypes of glioma patients, and could provide significant clinical auxiliary value for treatment decision-making, and prognosis assessment in glioma patients.

关键词

胶质瘤 / MRI影像组学 / 异柠檬酸脱氢酶 / 氧6-甲基鸟嘌呤-DNA甲基转移酶 / GA-fKPLS模型

Key words

Glioma / MRI radiomics / Isocitrate dehydrogenase / O6-methylguanine-DNA methyltransferase / GA-fKPLS model

引用本文

引用格式 ▾
杨琪,闫俊羽,刘夏,王彤,房瑞玲,曾平,杨海涛,罗艳虹,余红梅,张岩波,杨国强,曹红艳. 基于复合核偏最小二乘模型的胶质瘤多模态 MRI 数据预测 MGMT 启动子甲基化合并 IDH 突变*[J]. 中国卫生统计, 2026, 43(4): 482-488 DOI:10.11783/j.issn.1002-3674.2026.04.001

登录浏览全文

4963

注册一个新账户 忘记密码

参考文献

[1]

Ostrom QT, Bauchet L, Davis FG, et al. The epidemiology of glioma in adults: a "state of the science" review[J].Neuro—Oncology, 2014,16(7):896-913.

[2]

McNamara C, Mankad K, Thust S, et al. 2021 WHO classification of tumours of the central nervous system: a review for the neuroradiologist[J].Neuroradiology, 2022,64(10):1919-1950.

[3]

Itakura H, Achrol AS, Mitchell LA, et al. Magnetic resonance image features identify glioblastoma phenotypic subtypes with distinct molecular pathway activities[J].Science Translational Medicine, 2015,7(303):138-303.

[4]

Yang G, Sha Y, Wang X, et al. Radiomics Profiling Identifies the Incremental Value of MRI Features beyond Key Molecular Biomarkers for the Risk Stratification of High—Grade Gliomas[J].Contrast Media & Molecular Imaging, 2022(1):8952357.

[5]

Smits M, van den Bent MJ. Imaging correlates of adult glioma genotypes[J].Radiology, 2017,284(2):316-331.

[6]

Louis DN, Perry A, Wesseling P, et al. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary[J].Neuro—Oncology, 2021,23(8):1231-1251.

[7]

Molenaar RJ, Verbaan D, Lamba S, et al. The combination of IDH1 mutations and MGMT methylation status predicts survival in glioblastoma better than either IDH1 or MGMT alone[J].Neuro—Oncology, 2014,16(9):1263-1273.

[8]

Tanaka K, Sasayama T, Mizukawa K, et al. Combined IDH1 mutation and MGMT methylation status on long—term survival of patients with cerebral low—grade glioma[J].Clinical Neurology and Neurosurgery, 2015,138:37-44.

[9]

Leu S, von Felten S, Frank S, et al. IDH/MGMT—driven molecular classification of low—grade glioma is a strong predictor for long—term survival[J].Neuro—Oncology, 2013,15(4):469-479.

[10]

谢致岩. 扩散加权成像预测脑胶质瘤IDH 突变与MGMT 启动子甲基化共现的研究[D].青岛: 青岛大学, 2023.

[11]

Ibrahim A, Primakov S, Beuque M, et al. Radiomics for precision medicine: Current challenges, future prospects, and the proposal of a new framework[J].Methods, 2021,188:20-29.

[12]

Liu Z, Wang S, Dong D, et al. The Applications of Radiomics in Precision Diagnosis and Treatment of Oncology: Opportunities and Challenges[J].Theranostics, 2019,9(5):1303-1322.

[13]

Avanzo M, Wei L, Stancanello J, et al. Machine and deep learning methods for radiomics[J].Medical physics, 2020,47(5):e185-e202.

[14]

Song J, Yin Y, Wang H, et al. A review of original articles published in the emerging field of radiomics[J].European Journal of Radiology, 2020,127:108991.

[15]

沙永建, 王效春, 谭艳, 等. MRI影像组学术前预测较低级别胶质瘤IDH 突变合并MGMT 启动子甲基化亚型的价值[J].磁共振成像, 2022,13(7):6-11.

[16]

Tan Y, Zhang S, Wei J, et al. A radiomics nomogram may improve the prediction of IDH genotype for astrocytoma before surgery[J].European Radiology, 2019,29(7):3325-3337.

[17]

Kim M, Jung SY, Park JE, et al. Diffusion— and perfusion—weighted MRI radiomics model may predict isocitrate dehydrogenase (IDH) mutation and tumor aggressiveness in diffuse lower grade glioma[J].European Radiology, 2020,30(4):2142-2151.

[18]

Sun C, Fan L, Wang W, et al. Radiomics and qualitative features from multiparametric MRI predict molecular subtypes in patients with lower—grade glioma[J].Frontiers in Oncology, 2022, 11:756828.

[19]

Korfiatis P, Kline TL, Coufalova L, et al. MRI texture features as biomarkers to predict MGMT methylation status in glioblastomas[J].Medical physics, 2016,43(6Part1):2835-2844.

[20]

Yang H, Cao H, He T, et al. Multilevel heterogeneous omics data integration with kernel fusion[J].Briefings in Bioinformatics, 2020,21(1):156-170.

[21]

Meinshausen N, Bühlmann P. Stability selection[J].Journal of the Royal Statistical Society Series B: Statistical Methodology, 2010,72(4):417-473.

[22]

Rännar S, Lindgren F, Geladi P, et al. A PLS kernel algorithm for data sets with many variables and fewer objects. Part 1: Theory and algorithm[J].Journal of Chemometrics, 1994,8(2):111-125.

[23]

Huang X, Pan W. Linear regression and two—class classification with gene expression data[J].Bioinformatics, 2003,19(16):2072-2078.

[24]

Holland JH. Genetic algorithms[J].Scientific American, 1992,267(1):66-73.

[25]

Zhuo L, Zheng J, Li X, et al. A genetic algorithm based wrapper feature selection method for classification of hyperspectral images using support vector machine[C]//Geoinformatics 2008 and Joint Conference on GIS and Built Environment: Classification of Remote Sensing Images. Bellingham: SPIE, 2008.

[26]

夏志强,方羚,任凌云,等. 基于LASSO回归模型构建扩张型心肌病死亡风险的预测模型[J].中国卫生统计,2025,42(4):590-593.

[27]

AlOmair O, M Zakaria O, Alabdullatif M, et al. Optimized KNN with domain—informed features and LIME explainability for improved breast cancer classification[J].BMC Medical Informatics and Decision Making, 2026, 26(1): 138.

[28]

Stoltzfus JC. Logistic Regression: A Brief Primer[J].Academic Emergency Medicine, 2011,18(10):1099-1104.

[29]

Efron B. Bayes' Theorem in the 21st Century[J].Science, 2013,340(6137):1177-1178.

[30]

Wirsching H, Weller M. Glioblastoma[J].Malignant Brain Tumors: State—of—the—Art Treatment, 2017:265-288.

[31]

Zhang S, Sun H, Su X, et al. Automated machine learning to predict the co—occurrence of isocitrate dehydrogenase mutations and O6—methylguanine—DNA methyltransferase promoter methylation in patients with gliomas[J].Journal of Magnetic Resonance Imaging, 2021,54(1):197-205.

[32]

范廷勇, 李建彬, 于金明. 功能影像在脑胶质瘤放射治疗靶区确定中的价值[J].中华肿瘤杂志, 2006,28(10):721-723.

[33]

李方明, 聂青, 赵文锐, 等. CHO PET 功能显像:脑胶质瘤诊断和优化放化疗方案[J].中华神经外科疾病研究杂志, 2011,10(2):154-157.

基金资助

国家自然科学基金资助项目(82473739;82273742)

山西省基础研究计划(202303021211130)

山西省回国留学人员科研资助项目(2024-081)

山西省高等教育“百亿工程”科技引导专项(BYBLK003)

煤炭环境致病与防治教育部重点实验室(山西医科大学)开放课题基金(MEKLCEPP/SXMU-202415)

AI Summary AI Mindmap
PDF (1564KB)

0

访问

0

被引

详细

导航
相关文章

AI思维导图

/

〈 〉