Song Bin, Wang Hao, Chen Yongqi, Liu Weiyan, Wei Ran, Dai Zedong, Hu Wenjuan, Ding Yi, Wang Lanyun
Department of Radiology Department of Pathology Department of General Surgery, Minhang Branch, Zhongshan Hospital, Fudan University, Shanghai, China.
Medicine (Baltimore). 2018 Jun;97(26):e11279. doi: 10.1097/MD.0000000000011279.
To identify magnetic resonance imaging (MRI) features in the prediction of tumor aggressiveness in patients with papillary thyroid carcinoma (PTC).In this prospective study, 105 patients with 122 PTCs underwent MRI with T1-weighted, T2-weighted, diffusion-weighted imaging and contrast-enhanced sequences prior to thyroidectomy. Based on exclusion criteria, 62 patients with 62 PTCs were finally suitable for further analysis. Tumor aggressiveness was defined according to the surgical histopathology. Tumor size, apparent diffusion coefficients (ADC) value and MRI features on images were obtained for each patient. Descriptive statistics for tumor aggressiveness, sensitivity, specificity, and accuracy of individual features were determined. A multivariate logistic regression model was developed to identify features that were independently predictive for tumor aggressiveness. Analyses of receiver-operating characteristic (ROC) curve were performed.High aggressive PTC significantly differed from low aggressive PTC in size (P = .016), size classification (P < .001), ADC value (P = .01), angulation on the lateral surface of the lesion (P = .009), signal intensity heterogeneity on ADC maps (P = .003), early enhancement degree (P < .001), tumor margin on delayed contrast-enhanced images (P < .001), and inner lining of delayed ring enhancement (P = .028). The interobserver agreement between the 2 readers was satisfactory with Cohen k ranging from 0.83 to 1.00 (P < .001). Logistic regression model showed lesion size classification and tumor margin on delayed contrast-enhanced images as strongest independent predictors of high aggressive PTC (P = .009 and P = .047), with an accuracy of 83.9%. The area under ROC curve for ADC value and lesion size were 0.68 and 0.81, respectively.These findings suggest that MRI before surgery has the potential to discriminate tumor aggressiveness in patients with PTC.
识别甲状腺乳头状癌（PTC）患者肿瘤侵袭性预测中的磁共振成像（MRI）特征。在这项前瞻性研究中，105例患有122个PTC的患者在甲状腺切除术前接受了T1加权、T2加权、扩散加权成像和对比增强序列的MRI检查。根据排除标准，最终62例患有62个PTC的患者适合进一步分析。根据手术组织病理学定义肿瘤侵袭性。获取每位患者的肿瘤大小、表观扩散系数（ADC）值和图像上的MRI特征。确定肿瘤侵袭性、个体特征的敏感性、特异性和准确性的描述性统计量。建立多因素逻辑回归模型以识别对肿瘤侵袭性具有独立预测性的特征。进行受试者操作特征（ROC）曲线分析。高侵袭性PTC与低侵袭性PTC在大小（P = 0.016）、大小分类（P < 0.001）时、ADC值（P = 0.01）、病变外侧表面的角度（P = 0.009）、ADC图上的信号强度异质性（P = 0.003）、早期强化程度（P < 0.001）、延迟对比增强图像上的肿瘤边缘（P < 0.001）以及延迟环形强化的内层（P = 0.028）方面存在显著差异。两位阅片者之间的观察者间一致性令人满意，Cohen k值范围为0.83至1.00（P < 0.001）。逻辑回归模型显示病变大小分类和延迟对比增强图像上的肿瘤边缘是高侵袭性PTC的最强独立预测因素（P = 0.009和P = 0.047），准确率为83.9%。ADC值和病变大小的ROC曲线下面积分别为0.68和0.81。这些发现表明术前MRI有潜力区分PTC患者的肿瘤侵袭性。
