基于超声影像组学鉴别腮腺多形性腺瘤和腺淋巴瘤的临床应用研究  

Clinical application of ultrasonic radiomics in distinguishing pleomorphic adenoma and adenomatoma of parotid gland

作　　者：魏伟 冯慧俊 范莉芳 王晔 韦天军 张伟 张霞 Wei Wei;FENG Hui-jun;FAN Li-fang(Department of Ultrasound,the First Affiliated Hospital of Wannan Medical College(Yijishan Hospital),Anhui 241001,China)

机构地区：[1]皖南医学院第一附属医院(弋矶山医院)超声医学科,安徽芜湖241001 [2]皖南医学院医学影像学院 [3]皖南医学院第一附属医院(弋矶山医院)病理科,安徽芜湖241001

出　　处：《放射学实践》2024年第7期954-959,共6页Radiologic Practice

摘　　要：目的:建立并验证基于超声影像组学鉴别腮腺多形性腺瘤和腺淋巴瘤的临床应用价值。方法:回顾性收集本院2012年1月-2022年12月就诊于口腔颌面外科并手术后病理确诊为腮腺多形腺腺瘤和腺淋巴瘤的患者共282例,并获取完整的超声影像及临床资料。按7:3比例随机分为训练集(197例)和验证集(85例)。采用先组间差异性分析筛选出临床独立预测因子。使用ITK-SNAP对感兴趣区(ROI)勾画,利用Pyradiomics软件从腮腺肿瘤的超声图像中提取组学特征。利用Pearson相关性分析,保留相关性系数>0.9的一个特征;再利用最小收缩和选择算子(LASSO)回归分析,筛选出特征构建超声影像组学模型;与临床数据结合构建联合模型并评估其诊断效能。采用Delong检验评估各模型的鉴别效能。结果:研究发现,年龄、性别、回声不均能够有效鉴别腮腺多形性腺瘤和腺淋巴瘤,构建临床诊断模型。通过降维筛选出15个特征并建立超声影像组学模型。基于超声影像组学模型结合性别、年龄、回声不均匀构建联合诊断模型。在训练集中,联合模型的AUC均高于临床诊断模型和超声影像组学模型且差异均具有统计学意义(Z=3.919,P<0.001;Z=3.179,P=0.0015),超声影像组学模型的AUC与临床模型AUC差异无统计学意义(Z=0.079,P=0.936)。验证集中,联合模型AUC高于临床模型和超声影像组学模型,差异均具有统计学意义(Z=2.424,P=0.015;Z=2.212,P=0.027);超声影像组学模型的AUC高于临床模型,差异不具有统计学意义(Z=0.881,P=0.379)。结论:基于超声影像组学构建联合模型在腮腺多形性腺瘤和Warthin瘤具有良好的鉴别诊断效能。Objective:To establish and verify the clinical application value of ultrasound radiomics in differential diagnosis of pleomorphic adenoma and adenolymphoma of the parotid gland.Methods:Complete ultrasound images and clinical data of 282 patients with parotid pleomorphic adenoma and adenolymphoma diagnosed by postoperative pathology in the Department of Oral and Maxillofacial Surgery of our hospital from January 2012 to December 2022 were retrospectively collected analyzed.All cases were randomly divided into a training set(197 cases) and a validation set(85 cases) at a ratio of 7:3.The clinical independent predictors were screened by difference analysis between the two groups.ITK-SNAP was used to delineate the region of interest(ROI),and Pyradiomics software was used to extract radiomics features from ultrasound images of parotid tumors.The Pearson correlation analysis was used to retain a feature of correlation coefficient>0.9.Least Absolute Shrinkage and Selection Operator(LASSO) regression analysis was used to selected features to construct an ultrasonic radiomics model,which was then combined with clinical data to construct a joint model to evaluate its diagnostic efficacy.Delong test was used to evaluate the discriminative efficacy of each model.Results:It was found that age,gender and echo irregularity could effectively distinguish parotid pleomorphic adenoma from adenolymphoma,and a clinical diagnostic model was constructed.15 features were selected by dimensionality reduction and the ultrasound radiomics model was established.A combined diagnostic model was constructed based on this ultrasound radiomics model by combining with gender,age,and echo irregularity.In the training set,the AUC of the combined model was higher than that of the clinical diagnostic model and ultrasound radiomics model,and the differences were statistically significant(Z=3.919,P<0.001;Z=3.179,P=0.0015).There was no significant difference in AUC between the ultrasound radiomics model and the clinical model(Z=0.079,P=0.936).In validat

关 键 词：腮腺多形性腺瘤 腺淋巴瘤 超声 影像组学 诊断 鉴别 

分 类 号：R445.1[医药卫生—影像医学与核医学] R739.87[医药卫生—诊断学]