基于磁共振的放射组学预测高级别脑胶质瘤中 CDK6 的表达及预后价值。

MR-Based Radiomics Predicts CDK6 Expression and Prognostic Value in High-grade Glioma.

机构信息

Department of Neurosurgery, Beijing Tiantan Hospital, Capital Medical University, No. 119 West Road, South Fourth Ring Road, Beijing, China.

出版信息

Acad Radiol. 2024 Dec;31(12):5141-5153. doi: 10.1016/j.acra.2024.06.006. Epub 2024 Jul 4.

DOI:10.1016/j.acra.2024.06.006

PMID:38964985

Abstract

RATIONALE AND OBJECTIVES

This study aims to assess the prognostic value of Cyclin-dependent kinases 6 (CDK6) expression levels and establish a machine learning-based radiomics model for predicting the expression levels of CDK6 in high-grade gliomas (HGG).

MATERIALS AND METHODS

Clinical parameters and genomic data were extracted from 310 HGG patients in the Cancer Genome Atlas (TCGA) database and 27 patients in the Repository of Molecular Brain Neoplasia Data (REMBRANDT) database. Univariate and multivariate Cox regression, as well as Kaplan-Meier analysis, were performed for prognosis analysis. The correlation between immune cell Infiltration with CDK6 was assessed using spearman correlation analysis. Radiomic features were extracted from contrast-enhanced magnetic resonance imaging (CE-MRI) in the Cancer Imaging Archive (TCIA) database (n = 82) and REMBRANDT database (n = 27). Logistic regression (LR) and support vector machine (SVM) were employed to establish the radiomics model for predicting CDK6 expression. Receiver operating characteristic curve (ROC), calibration curve, and decision curve analysis (DCA) were utilized to assess the predictive performance of the radiomics model. Generate radiomic scores (RS) based on the LR model. An RS-based nomogram was constructed to predict the prognosis of HGG.

RESULTS

CDK6 was significantly overexpressed in HGG tissues and was related to lower overall survival. A significant elevation in infiltrating M0 macrophages was observed in the CDK6 high group (P < 0.001). The LR radiomics model for the prediction of CDK6 expression levels (AUC=0.810 in the training cohort, AUC = 0.784 after cross-validation, AUC=0.750 in the testing cohort) was established utilizing three radiomic features. The predictive efficiencies of the RS-based nomogram, as measured by AUC, were 0.769 for 1-year, 0.815 for 3-year, and 0.780 for 5-year, respectively.

CONCLUSION

The expression level of CDK6 can impact the prognosis of patients with HGG. The expression level of HGG can be noninvasively prognosticated utilizing a radiomics model.

摘要

背景和目的

本研究旨在评估细胞周期蛋白依赖性激酶 6（CDK6）表达水平的预后价值，并建立基于机器学习的放射组学模型，以预测高级别胶质瘤（HGG）中 CDK6 的表达水平。

材料和方法

从癌症基因组图谱（TCGA）数据库的 310 例 HGG 患者和 REMBRANDT 数据库的 27 例患者中提取临床参数和基因组数据。进行单变量和多变量 Cox 回归以及 Kaplan-Meier 分析以进行预后分析。使用 Spearman 相关分析评估 CDK6 与免疫细胞浸润的相关性。从癌症成像档案（TCIA）数据库（n=82）和 REMBRANDT 数据库（n=27）的对比增强磁共振成像（CE-MRI）中提取放射组学特征。采用逻辑回归（LR）和支持向量机（SVM）建立预测 CDK6 表达的放射组学模型。利用受试者工作特征曲线（ROC）、校准曲线和决策曲线分析（DCA）评估放射组学模型的预测性能。基于 LR 模型生成放射组学评分（RS）。构建基于 RS 的列线图以预测 HGG 的预后。

结果

CDK6 在 HGG 组织中明显过表达，与总体生存率降低相关。在 CDK6 高表达组中，浸润性 M0 巨噬细胞显著升高（P<0.001）。利用三个放射组学特征建立了预测 CDK6 表达水平的 LR 放射组学模型（训练队列的 AUC=0.810，交叉验证后的 AUC=0.784，测试队列的 AUC=0.750）。基于 RS 的列线图的预测效率，通过 AUC 测量，分别为 1 年的 0.769、3 年的 0.815 和 5 年的 0.780。