| 摘要: |
| 目的：探讨超声影像组学对腮腺多形性腺瘤及腺淋巴瘤鉴别诊断能力。方法：回顾性收集2016年1月-2021年9月于皖南医学院第一附属医院（弋矶山医院）经手术病理证实的133例多形性腺瘤和99例腺淋巴瘤的超声图像及临床相关资料。按7∶3的比例分为训练集162例、验证集70例。利用ITK-SNAP软件上手动勾画肿瘤的感兴趣区（Region Of Interest，ROI），采用Pyradiomics 软件提取二维超声腮腺病灶影像组学定量特征，将提取特征正则化，采用Spearman相关分析，再使用选择最小绝对收缩与选择算子（Least Absolute Shrinkage and Selection Operator，LASSO）模型，筛选出最佳特征。分别采用支持向量机（SVM）、K紧邻（KNN）、决策树（Decision tree）三种机器学习算法根据筛选的最佳特征构建超声影像组学模型，建立对多形性腺瘤和腺淋巴瘤鉴别诊断能力。绘制受试者操作特征（Receiver Operating Characteristic，ROC）曲线评价鉴别各个模型诊断的效能。采用Delong检验评估模型的差异，应用决策曲线分析（DCA）评估模型的临床应用价值。结果：超声影像组学鉴别腮腺常见的良性肿瘤时，提取影像组学21个特征，利用SVM、KNN、Decision tree算法所构建模型鉴别效能AUC、灵敏度、特异度、准确度分别是：0.848、0.842、0.814、0.826；0.721、0.947、0.481、0.652；0.620、0.684、1.000、0.608。通过DeLong检验发现SVM算法优于其他两种算法所构建的模型。结论：超声影像组学利用机器算法可以用于多形性腺瘤与腺淋巴瘤的鉴别。 |
| 关键词: 超声影像组学 多形性腺瘤 腺淋巴瘤 机器算法 |
| DOI： |
| 投稿时间：2023-02-23修订日期：2023-06-13 |
| 基金项目:皖南医学院中青年科研基金(编号:WK2020F02) |
|
| To investigate the value of ultrasound radiomics in differential diagnosis of parotid pleomorphic adenoma and adenolymphoma |
| Feng huijun,Zhu huiling,Jiang feng |
| () |
| Abstract: |
| Objective: To investigate the ability of ultrasound radiomics in the differential diagnosis of pleomorphic adenoma and adenolymphoma lymphoma in parotid gland. Methods: The ultrasonic images and clinical data of 133 cases of pleomorphic adenoma and 99 cases of adenomas confirmed by postoperative pathology were retrospectively collected from the First Affiliated Hospital of Wannan Medical College (Yijishan Hospital) from January 2016 to September 2021. According to the ratio of 7∶3, there were 162 cases in training set and 70 cases in verification set.The Region Of Interest (ROI) of the tumor was manually delineated on ITK-SNAP software, and the quantitative features of two-dimensional ultrasonic parotid focus imaging were extracted by Pyradiomics software. The extracted features were regularized and Spearman correlation analysis was performed. Then, the model of Least Absolute Shrinkage and Selection Operator (LASSO) can be used to screen out the best features. Three machine learning algorithms, namely support vector machine (SVM), K-Nearest Neighbor (KNN) and Decision tree, were used to construct the characteristics of ultrasonic image omics model, and establish the differential diagnosis ability of pleomorphic adenoma and adenoma. Receiver Operating Characteristic (ROC) curves were plotted to evaluate the diagnostic efficiency of each model. Delong test was used to evaluate the differences of the models, and decision curve analysis (DCA) was used to evaluate the clinical application value of the models. Results: When ultrasound radiomics was used to identify common benign tumors in parotids, 21 radiomics features were extracted. The AUC, sensitivity, specificity and accuracy of the model constructed by SVM, KNN and Decision Tree algorithm were 0.848, 0.842, 0.814 and 0.826, respectively. 0.721, 0.947, 0.481, 0.652; 0.620, 0.684, 1.000, 0.608. Through DeLong test, it is found that SVM algorithm is superior to the model constructed by other two algorithms. Conclusion: Ultrasound radiomics can be used to distinguish pleomorphic adenoma from adenolymphoma by machine algorithm. |
| Key words: Ultrasound Radiomics Parotid Pleomorphic Adenoma Parotid Adenolymphoma Machine Algorithm |
