Department of Radiology, Northwestern Memorial Hospital, Northwestern University Feinberg School of Medicine, 676 N. St. Clair St. Suite 800, Chicago, IL, 60611, USA.
Department of Anesthesiology, Rush University, Chicago, IL, 60612, USA.
Abdom Radiol (NY). 2020 Sep;45(9):2797-2809. doi: 10.1007/s00261-020-02637-w. Epub 2020 Jul 14.
Solid renal masses (SRM) are difficult to differentiate based on standard MR features. The purpose of this study was to assess MR imaging features of SRM to evaluate performance of ensemble methods of classifying SRM subtypes.
MR images of SRM (n = 330) were retrospectively evaluated for standard and multiparametric (mp) features. Models of MR features for predicting malignant and benign lesions as well as subtyping SRM were developed using a training dataset and performance was evaluated in a test data-set using recursive partitioning (RP), gradient booting machine (GBM), and random forest (RF) methods.
In the test dataset, GBM and RF models demonstrated an accuracy of 86% (95% CI 75% to 93%) for predicting benign versus malignant SRM compared to 83% (95% CI 71% to 91%) for the RP model. RF had the greatest accuracy in predicting SRM subtypes, 81.2% (95% CI 69.5% to 89.9%) compared with GBM 73.4% (95% CI 60.9% to 83.7%) or RP 70.3% (95% CI 57.6% to 81.1%). Marginal homogeneity was reduced by the RF model compared with the RP model (P < 0.001), but not the GBM model (P = 0.135). All models had high sensitivity and specificity for clear cell and papillary renal cell carcinomas (RCC), but performed less well in differentiating chromophobe RCC, oncocytomas, and fat-poor angiomyolipomas.
Ensemble methods for prediction of SRM from radiologist-assessed image characteristics have high accuracy for distinguishing benign and malignant lesions. SRM subtype classification is limited by the ability to categorize chromophobe RCCs, oncocytomas, and fat-poor angiomyolipomas.
基于标准磁共振（MR）特征，很难对实体性肾脏肿块（SRM）进行区分。本研究旨在评估 SRM 的 MR 成像特征，以评估分类 SRM 亚型的集成方法的性能。
对 330 例 SRM 的 MR 图像进行回顾性评估，以评估标准和多参数（mp）特征。使用训练数据集为预测良恶性病变以及 SRM 亚型分类建立基于 MR 特征的模型，并使用递归分区（RP）、梯度提升机（GBM）和随机森林（RF）方法在测试数据集上评估性能。
在测试数据集中，GBM 和 RF 模型预测良恶性 SRM 的准确率为 86%（95%CI，75%93%），而 RP 模型的准确率为 83%（95%CI，71%91%）。RF 模型在预测 SRM 亚型方面具有最高的准确性，为 81.2%（95%CI，69.5%89.9%），而 GBM 为 73.4%（95%CI，60.9%83.7%），RP 为 70.3%（95%CI，57.6%~81.1%）。与 RP 模型相比，RF 模型降低了边缘均匀性（P<0.001），但与 GBM 模型相比差异无统计学意义（P=0.135）。所有模型对透明细胞和乳头状肾细胞癌（RCC）均具有较高的敏感性和特异性，但在区分嫌色细胞 RCC、嗜酸细胞瘤和乏脂性血管平滑肌脂肪瘤方面效果较差。
基于放射科医生评估的图像特征预测 SRM 的集成方法对区分良恶性病变具有较高的准确性。SRM 亚型分类受到将嫌色细胞 RCC、嗜酸细胞瘤和乏脂性血管平滑肌脂肪瘤进行分类的能力限制。
