Department of Ultrasound, The Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, 310000, China.
Dalian University of Technology, Dalian, 116024, China.
Cancer Lett. 2022 Feb 28;527:107-114. doi: 10.1016/j.canlet.2021.12.015. Epub 2021 Dec 17.
Although conventional ultrasound (CUS) allows for clear detection of parotid gland lesions (PGLs), it fails to accurately provide benign-malignant differentiation due to overlapping morphological features. Radiomics is capable of processing large-quantity volume of data hidden in CUS image undiscovered by naked eyes. The aim was to explore the potential of CUS-based radiomics score (Rad-score) in distinguishing benign (BPGLs) and malignant PGLs (MPGLs). A consecutive of 281 PGLs (197 in training set and 84 in test set) with definite pathological confirmation was retrospectively enrolled. 1465 radiomics features were extracted from CUS images and Rad-score was constructed by using Least Absolute Shrinkage and Selection Operator (LASSO) algorithm. Different nomogram models, including clinic-radiomics (Clin + Rad-score), CUS-clinic (CUS + Clin) and combined CUS-clinic-radiomics (CUS + Clin + Rad-score), were built using logistic regression. The diagnostic performance of different models were calculated and compared by area under receiver operating curve (AUC) and corresponding sensitivity and specificity. Finally, 26 radiomics features were independent signatures for predicting MPGLs, with MPGLs having higher Rad-scores in both cohorts (both P < 0.05). In the test population, CUS + Clin + Rad-score obtained an excellent diagnostic result, with significantly higher AUC value (AUC = 0.91) when compared to that of CUS + Clin (AUC = 0.84) and Clin + Rad-score (AUC = 0.74), respectively (both P < 0.05). In addition, the sensitivity of this combined model was higher than that of single Rad-score model (100.00% vs. 71.43%, P = 0.031) without compromising the specificity value (82.86% vs. 88.57%, P = 0.334). The calibration curve and decision curve analysis also indicated the clinical effectiveness of the proposed combined nomogram. The combined CUS-clinic-radiomics model may help improve the discrimination of BPGLs from MPGLs.
虽然常规超声（CUS）能够清晰地检测腮腺病变（PGLs），但由于形态学特征重叠，无法准确进行良恶性鉴别。放射组学能够处理大量隐藏在 CUS 图像中肉眼无法发现的数据。本研究旨在探讨基于 CUS 的放射组学评分（Rad-score）在鉴别良性（BPGLs）和恶性 PGLs（MPGLs）中的潜力。回顾性纳入了 281 例经病理证实的 PGLs（训练组 197 例，测试组 84 例）。从 CUS 图像中提取 1465 个放射组学特征，采用最小绝对收缩和选择算子（LASSO）算法构建 Rad-score。使用逻辑回归构建不同的列线图模型，包括临床-放射组学（Clin+Rad-score）、CUS-临床（CUS+Clin）和联合 CUS-临床-放射组学（CUS+Clin+Rad-score）。通过计算受试者工作特征曲线（ROC）下面积（AUC）及其对应的敏感性和特异性来评估不同模型的诊断性能。最后，筛选出 26 个与预测 MPGLs 相关的独立放射组学特征，两组队列中 MPGLs 的 Rad-score 均较高（均 P<0.05）。在测试人群中，CUS+Clin+Rad-score 获得了出色的诊断结果，与 CUS+Clin（AUC=0.84）和 Clin+Rad-score（AUC=0.74）相比，AUC 值显著更高（AUC=0.91，均 P<0.05）。此外，与单独的 Rad-score 模型相比，该联合模型的敏感性更高（100.00%比 71.43%，P=0.031），而特异性值无差异（82.86%比 88.57%，P=0.334）。校准曲线和决策曲线分析也表明了所提出的联合列线图的临床有效性。联合 CUS-临床-放射组学模型可能有助于提高 BPGLs 与 MPGLs 的鉴别能力。
