基于MRI影像组学模型在颅内生殖细胞瘤和颅咽管瘤诊断的应用研究  

Application of MRI Radiomics Model to the Diagnosisof Intracranial Germinoma and Craniopharyngioma

作　　者：龚江年 黄丽轩 黎丽 梁伦 曾自三[1] GONG Jiangnian;HUANG Lixuan(Department of Radiology,The First Affiliated Hospital of Guangxi Medical University,Nanning,Guangxi Province 530021,P.R.China)

机构地区：[1]广西医科大学第一附属医院医学影像科,南宁530021 [2]广西医科大学第一附属医院神经外科,南宁530021

出　　处：《临床放射学杂志》2025年第3期411-416,共6页Journal of Clinical Radiology

基　　金：国家自然科学基金地区科学基金项目(编号:82460337);广西重点研发计划(编号:桂科AB24010172);广西自然科学基金区域高发疾病研究联合专项资助(编号:2024GXNSFAA010315)。

摘　　要：目的探索基于MRI影像组学特征的预测模型在颅内生殖细胞瘤(IG)和颅咽管瘤(CP)中的鉴别价值。方法搜集了114例(50例颅内生殖细胞瘤,64例颅咽管瘤)患者的MRI扫描图像,在常规轴位T 2WI和T 2FLAIR图像上勾画肿瘤实性强化部分为感兴趣区(ROI)。通过合并IG和CP的类别信息,对T 2WI和T 2FLAIR序列的影像组学特征分别进行提取及融合,采用Pearson相关系数、最大相关最小冗余(mRMR)及最小绝对收缩和选择算子(LASSO)进行影像组学特性选择,运用逻辑回归(LR)分类器来构建单序列影像组学特征模型及两序列联合影像组学特征模型,对训练集进行5次交叉验证,运用受试者工作特征曲线曲线下面积、准确性、敏感度、特异度、阳性预测值、阴性预测值和95%可信区间对预测模型的预测效能进行评价。结果从T 2WI和T 2FLAIR序列中提取的图像特征中筛选了26个最具有代表性的特征。其中T 2WI和T 2FLAIR序列联合模型具有更显著的预测性能,它的受试者工作特征曲线曲线下面积、准确率、敏感度和特异度分别为0.874、77.1%、78.6%和76.2%,表现出良好的鉴别性能。结论基于T 2WI和T 2FLAIR影像组学特征的联合模型可以更好地实现对IG和CP的鉴别价值并为临床决策提供了一种无创和有效的方法。Objective To explore the diagnostic value of a prediction model based on MRI radiomics features in differentiating intracranial germinoma(IG)from craniopharyngioma(CP).Methods MRI scans of 114 patients(50 with intracranial germinomas and 64 with craniopharyngiomas)were collected.The region of interest(ROI)of the solid enhancement part of the tumor was delineated on conventional axial T 2-weighted imaging(T 2WI)and T 2 fluid-attenuated inversion recovery(T 2FLAIR)images.Radiomics features were extracted and fused from T 2WI and T 2FLAIR sequences after combining the category information of IG and CP.Radiomics features were selected using Pearson correlation coefficients,minimal redundancy maximal relevance(mRMR),and least absolute shrinkage and selection operator(LASSO).Logistic regression(LR)classifiers were used to construct single-sequence and combined-sequence radiomics feature models.The training set was cross-validated five times,and the prediction performance of the models was evaluated using the area under the receiver operating characteristic curve(AUC),accuracy,sensitivity,specificity,positive predictive value,negative predictive value,and 95%confidence interval.Results A total of 26 most representative features were selected from radiomics features extracted from T 2WI and T 2FLAIR sequences.The combined model of T 2WI and T 2FLAIR sequences demonstrated superior prediction performance,achieving an AUC of 0.874,accuracy of 77.1%,sensitivity of 78.6%,and specificity of 76.2%.Conclusion The combined model based on T 2WI and T 2FLAIR radiomics features effectively differentiates IG from CP,offering a non-invasive and reliable method for clinical decision-making.

关 键 词：生殖细胞瘤 颅咽管瘤 磁共振成像 影像组学 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]