Department of Stomatology, Peking Union Medical College Hospital, Peking Union Medical College, Chinese Academy of Medical Sciences, Beijing, China.
Department of Implant Dentistry, Beijing Stomatological Hospital, Capital Medical University, Beijing, China.
Oral Surg Oral Med Oral Pathol Oral Radiol. 2021 May;131(5):591-599. doi: 10.1016/j.oooo.2021.01.014. Epub 2021 Jan 15.
The objective of this study was to compare the diagnostic performance of magnetic resonance imaging (MRI) and computed tomography (CT) in differentiating pleomorphic adenomas from Warthin tumors using radiomics.
We retrospectively reviewed 626 patients who underwent preoperative MRI or CT for parotid tumor diagnosis. Patient groups were balanced by propensity score matching (PSM) and 123 radiomic features were extracted from tumor images. Radiomic signatures (rad-scores) were generated using a least absolute shrinkage and selection operator logistic regression model. The Canny edge detector was used to define tumor borders (border index). The diagnostic performance of rad-score and border index before and after PSM was evaluated with area under the receiver operating characteristic curve analysis.
For differentiation of pleomorphic adenomas and Warthin tumors, rad-score and border index areas under the curve for MRI after PSM were 0.911 (95% confidence interval [CI], 0.871-0.951) and 0.716 (95% CI, 0.646-0.787), respectively; those for CT were 0.876 (95% CI, 0.829-0.923) and 0.608 (95% CI, 0.527-0.690), respectively. Tumor border index on MRI, but not CT, had superior diagnostic performance (P < .05); MRI- and CT-based rad-scores showed similar performance (P >.05).
MRI is superior to CT for tumor margin examination; however, the radiomics features of both modalities showed no difference.
本研究旨在通过放射组学比较磁共振成像(MRI)和计算机断层扫描(CT)在鉴别多形性腺瘤和沃辛瘤中的诊断性能。
我们回顾性分析了 626 例因腮腺肿瘤行术前 MRI 或 CT 检查的患者。采用倾向评分匹配(PSM)平衡患者组,从肿瘤图像中提取 123 个放射组学特征。使用最小绝对值收缩和选择算子逻辑回归模型生成放射组学特征(rad-score)。使用 Canny 边缘探测器定义肿瘤边界(边界指数)。使用接受者操作特征曲线下面积分析评估 PSM 前后 rad-score 和边界指数的诊断性能。
对于多形性腺瘤和沃辛瘤的鉴别,PSM 后 MRI 的 rad-score 和边界指数的曲线下面积分别为 0.911(95%置信区间[CI],0.871-0.951)和 0.716(95%CI,0.646-0.787);CT 分别为 0.876(95%CI,0.829-0.923)和 0.608(95%CI,0.527-0.690)。MRI 上的肿瘤边界指数具有更好的诊断性能(P<0.05),但 CT 上的肿瘤边界指数没有(P>0.05);MRI 和 CT 基于的 rad-score 表现相似(P>0.05)。
MRI 优于 CT 检查肿瘤边界;然而,两种方式的放射组学特征无差异。