Shen Nanxi, Lv Wenzhi, Li Shihui, Liu Dong, Xie Yan, Zhang Ju, Zhang Jiaxuan, Jiang Jingjing, Jiang Rifeng, Zhu Wenzhen
Department of Radiology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China.
Department of Artificial Intelligence, Julei Technology Company, Wuhan, China.
J Magn Reson Imaging. 2023 Mar;57(3):884-896. doi: 10.1002/jmri.28378. Epub 2022 Aug 5.
Noninvasive determination of Notch signaling is important for prognostic evaluation and therapeutic intervention in glioma.
To predict Notch signaling using multiparametric (mp) MRI radiomics and correlate with biological characteristics in gliomas.
Retrospective.
A total of 63 patients for model construction and 47 patients from two public databases for external testing.
FIELD STRENGTH/SEQUENCE: A 1.5 T and 3.0 T, T1-weighted imaging (T1WI), T2WI, T2 fluid attenuated inversion recovery (FLAIR), contrast-enhanced (CE)-T1WI.
Radiomic features were extracted from CE-T1WI, T1WI, T2WI, and T2FLAIR and imaging signatures were selected using a least absolute shrinkage and selection operator. Diagnostic performance was compared between single modality and a combined mpMRI radiomics model. A radiomic-clinical nomogram was constructed incorporating the mpMRI radiomic signature and Karnofsky Performance score. The performance was validated in the test set. The radiomic signatures were correlated with immunohistochemistry (IHC) analysis of downstream Notch pathway components.
Receiver operating characteristic curve, decision curve analysis (DCA), Pearson correlation, and Hosmer-Lemeshow test. A P value < 0.05 was considered statistically significant.
The radiomic signature derived from the combination of all sequences numerically showed highest area under the curve (AUC) in both training and external test sets (AUCs of 0.857 and 0.823). The radiomics nomogram that incorporated the mpMRI radiomic signature and KPS status resulted in AUCs of 0.891 and 0.859 in the training and test sets. The calibration curves showed good agreement between prediction and observation in both sets (P= 0.279 and 0.170, respectively). DCA confirmed the clinical usefulness of the nomogram. IHC identified Notch pathway inactivation and the expression levels of Hes1 correlated with higher combined radiomic scores (r = -0.711) in Notch1 mutant tumors.
The mpMRI-based radiomics nomogram may reflect the intratumor heterogeneity associated with downstream biofunction that predicts Notch signaling in a noninvasive manner.
3 TECHNICAL EFFICACY: Stage 2.
无创测定Notch信号对于胶质瘤的预后评估和治疗干预具有重要意义。
利用多参数（mp）MRI影像组学预测Notch信号，并与胶质瘤的生物学特征相关联。
回顾性研究。
共63例患者用于模型构建，47例患者来自两个公共数据库用于外部验证。
场强/序列：1.5T和3.0T，T1加权成像（T1WI）、T2WI、T2液体衰减反转恢复序列（FLAIR）、对比增强（CE）-T1WI。
从CE-T1WI、T1WI、T2WI和T2FLAIR中提取影像组学特征，并使用最小绝对收缩和选择算子选择影像特征。比较单模态与联合mpMRI影像组学模型的诊断性能。构建包含mpMRI影像特征和卡诺夫斯基性能评分的影像组学-临床列线图。在测试集中验证其性能。将影像特征与Notch信号通路下游成分的免疫组织化学（IHC）分析相关联。
受试者操作特征曲线、决策曲线分析（DCA）、Pearson相关性分析和Hosmer-Lemeshow检验。P值<0.05被认为具有统计学意义。
在训练集和外部测试集中，由所有序列组合得出的影像特征在数值上均显示出最高的曲线下面积（AUC）（AUC分别为0.857和0.823）。纳入mpMRI影像特征和KPS状态的影像组学列线图在训练集和测试集中的AUC分别为0.891和0.859。校准曲线显示两组预测值与观察值之间具有良好的一致性（P值分别为0.279和0.170）。DCA证实了列线图的临床实用性。IHC鉴定出Notch信号通路失活，且在Notch1突变型肿瘤中，Hes1的表达水平与较高的联合影像组学评分相关（r = -0.711）。
基于mpMRI的影像组学列线图可能反映与下游生物功能相关的肿瘤内异质性，以无创方式预测Notch信号。
3级 技术效能：2级
