Sun Shuangshuang, Li Lin, Xu Mengying, Wei Ying, Shi Feng, Liu Song
Department of Radiology, Nanjing Drum Tower Hospital, Affiliated Hospital of Medical School, Nanjing University, No. 321 Zhongshan Road, Nanjing, 210008, Jiangsu, China.
Department of Pathology, Nanjing Drum Tower Hospital, Affiliated Hospital of Medical School, Nanjing University, Nanjing, 210008, China.
Abdom Radiol (NY). 2024 Jun;49(6):1779-1791. doi: 10.1007/s00261-024-04306-8. Epub 2024 Apr 24.
To analyze the clinicopathologic information and CT imaging features of Epstein-Barr virus (EBV)-positive gastric cancer (GC) and establish CT-based radiomics models to predict the EBV status of GC.
This retrospective study included 144 GC cases, including 48 EBV-positive cases. Pathological and immunohistochemical information was collected. CT enlarged LN and morphological characteristics were also assessed. Radiomics models were constructed to predict the EBV status, including decision tree (DT), logistic regression (LR), random forest (RF), and support vector machine (SVM).
T stage, Lauren classification, histological differentiation, nerve invasion, VEGFR2, E-cadherin, PD-L1, and Ki67 differed significantly between the EBV-positive and -negative groups (p = 0.015, 0.030, 0.006, 0.022, 0.028, 0.030, < 0.001, and < 0.001, respectively). CT enlarged LN and large ulceration differed significantly between the two groups (p = 0.019 and 0.043, respectively). The number of patients in the training and validation cohorts was 100 (with 33 EBV-positive cases) and 44 (with 15 EBV-positive cases). In the training cohort, the radiomics models using DT, LR, RF, and SVM yielded areas under the curve (AUCs) of 0.905, 0.771, 0.836, and 0.886, respectively. In the validation cohort, the diagnostic efficacy of radiomics models using the four classifiers were 0.737, 0.722, 0.751, and 0.713, respectively.
A significantly higher proportion of CT enlarged LN and a significantly lower proportion of large ulceration were found in EBV-positive GC. The prediction efficiency of radiomics models with different classifiers to predict EBV status in GC was good.
分析爱泼斯坦-巴尔病毒（EBV）阳性胃癌（GC）的临床病理信息和CT影像特征，并建立基于CT的影像组学模型以预测GC的EBV状态。
这项回顾性研究纳入了144例GC病例，其中48例为EBV阳性病例。收集了病理和免疫组化信息。还评估了CT上肿大淋巴结及形态学特征。构建了用于预测EBV状态的影像组学模型，包括决策树（DT）、逻辑回归（LR）、随机森林（RF）和支持向量机（SVM）。
EBV阳性组和阴性组在T分期、劳伦分类、组织学分化、神经侵犯、VEGFR2、E-钙黏蛋白、PD-L1和Ki67方面存在显著差异（p值分别为0.015、0.030、0.006、0.022、0.028、0.030 <0.001和<0.001）。两组在CT上肿大淋巴结和大溃疡方面存在显著差异（p值分别为0.019和0.043）。训练队列和验证队列中的患者数量分别为100例（其中33例EBV阳性病例）和44例（其中15例EBV阳性病例）。在训练队列中，使用DT、LR、RF和SVM的影像组学模型的曲线下面积（AUC）分别为0.905、0.771、0.836和0.886。在验证队列中，使用这四种分类器的影像组学模型的诊断效能分别为0.737、0.722、0.751和0.713。
EBV阳性GC中CT上肿大淋巴结的比例显著更高，大溃疡的比例显著更低。不同分类器的影像组学模型预测GC中EBV状态的效率良好。
