Ma Zelan, Fang Mengjie, Huang Yanqi, He Lan, Chen Xin, Liang Cuishan, Huang Xiaomei, Cheng Zixuan, Dong Di, Liang Changhong, Xie Jiajun, Tian Jie, Liu Zaiyi
Department of Radiology, Guangdong General Hospital, Guangdong Academy of Medical Sciences, Guangzhou, 510080, China.
Key Laboratory of Molecular Imaging, Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China; University of Chinese Academy of Sciences, Beijing, 100190, China.
Eur J Radiol. 2017 Jun;91:142-147. doi: 10.1016/j.ejrad.2017.04.007. Epub 2017 Apr 12.
To evaluate the value of CT-based radiomics signature for differentiating Borrmann type IV gastric cancer (GC) from primary gastric lymphoma (PGL).
40 patients with Borrmann type IV GC and 30 patients with PGL were retrospectively recruited. 485 radiomics features were extracted and selected from the portal venous CT images to build a radiomics signature. Subjective CT findings, including gastric wall peristalsis, perigastric fat infiltration, lymphadenopathy below the renal hila and enhancement pattern, were assessed to construct a subjective findings model. The radiomics signature, subjective CT findings, age and gender were integrated into a combined model by multivariate analysis. The diagnostic performance of these three models was assessed with receiver operating characteristics curves (ROC) and were compared using DeLong test.
The subjective findings model, the radiomics signature and the combined model showed a diagnostic accuracy of 81.43% (AUC [area under the curve], 0.806; 95% CI [confidence interval]: 0.696-0.917; sensitivity, 63.33%; specificity, 95.00%), 84.29% (AUC, 0.886 [95% CI: 0.809-0.963]; sensitivity, 86.67%; specificity, 82.50%), 87.14% (AUC, 0.903 [95%CI: 0.831-0.975]; sensitivity, 70.00%; specificity, 100%), respectively. There were no significant differences in AUC among these three models (P=0.051-0.422).
Radiomics analysis has the potential to accurately differentiate Borrmann type IV GC from PGL.
评估基于CT的影像组学特征对鉴别Borrmann IV型胃癌（GC）与原发性胃淋巴瘤（PGL）的价值。
回顾性纳入40例Borrmann IV型GC患者和30例PGL患者。从门静脉期CT图像中提取并选择485个影像组学特征以构建影像组学特征模型。评估主观CT表现，包括胃壁蠕动、胃周脂肪浸润、肾门以下淋巴结肿大及强化方式，以构建主观表现模型。通过多变量分析将影像组学特征、主观CT表现、年龄和性别整合为一个联合模型。使用受试者操作特征曲线（ROC）评估这三种模型的诊断性能，并采用DeLong检验进行比较。
主观表现模型、影像组学特征模型和联合模型的诊断准确率分别为81.43%（曲线下面积[AUC]，0.806；95%置信区间[CI]：0.696 - 0.917；敏感性，63.33%；特异性，95.00%）、84.29%（AUC，0.886[95%CI：0.809 - 0.963]；敏感性，86.67%；特异性，82.50%）、87.14%（AUC，0.903[95%CI：0.831 - 0.975]；敏感性，70.00%；特异性，100%）。这三种模型的AUC之间无显著差异（P = 0.051 - 0.422）。
影像组学分析有潜力准确鉴别Borrmann IV型GC与PGL。
