伴有囊变的腮腺病变预测模型的建立与验证  

Construction and Verification of Predicting Models on Parotid Gland Lesions with Intratumorally Cystic Components

作　　者：易芹芹 周宙 凌人男 YI Qin-qin;ZHOU Zhou;LING Ren-nan(Department of Radiology,Shenzhen People’s Hospital,Shenzhen 508001,Guangdong Province,China)

机构地区：[1]深圳市人民医院放射科,广东深圳518001

出　　处：《中国CT和MRI杂志》2023年第1期44-46,共3页Chinese Journal of CT and MRI

基　　金：广东省医学科学技术研究基金项目(B2020004)。

摘　　要：目的 筛选并分析影响伴有囊变的腮腺病变良恶性诊断及多形性腺瘤/腺淋巴瘤诊断的因素,建立预测模型并验证。方法 回顾性分析2018年1月至2020年3月我院伴有囊变的腮腺良恶性病变共131例,搜集临床及CT特征共13项;92例作为训练组(A组),39例作为验证组(B组);另取其中多形性腺瘤及腺淋巴瘤共83例,58例作为训练组(C组),25例作为验证组(D组);通过单因素分析及Logistic回归分析A组及C组资料,筛选出独立影响因子,建立回归方程,将B组及D组资料分别代入方程进行验证。结果 A组Logistic回归分析示边界不清为恶性病变的独立影响因素,建立的回归方程训练组AUC为0.932,验证组AUC为0.885;C组Logistic回归分析示囊变不位于中央、男性及具有贴边血管征为腺淋巴瘤的独立影响因素,建立的回归方程训练组AUC为0.896,验证组AUC为0.880。结论 对于伴有囊变的腮腺病变,CT增强可以鉴别良恶性病变及腺淋巴瘤/多形性腺瘤,据此建立的预测模型具有良好的诊断效能。Objective To construct and verify predicting models using multivariate logistic regression analysis for estimating the probability of malignancy or adenolymphoma in parotid gland lesions with intratumorally cystic components. Methods From January 2018 to March 2017, 131 patients with parotid gland lesions accompanying intratumorally cystic change were analyzed retrospectively.Thirteen characteristics were collected. Among 131 cases, 92 patients were in modeling group(group A), and 39 patients were in validating group(group B). A total of 83 patients of pleomorphic adenoma and adenolymphoma were included in this study, which were also classified into two groups:58 patients were in modeling group(group C), and25 patients were in validating group(group D). Independent influencing factors of group A and group C by univariate analysis and multivariate logistic regression analysis were chosen and two predicting models of benign/malignant lesions and pleomorphic adenoma/adenolymphoma were established. Two models were verified with the date of group B and group D. Results Unclear border was an independent influencing factor in multivariate logistic regression analysis of group A. Centrally/eccentrically located cystic lesions, gender and peripheral vascular sign were independent influencing factors in multivariate logistic regression analysis of group C. Two predicting models were built basing on logistic regression analysis. AUC were 0.932 and 0.896 in modeling groups respectively. When group B and group D were substituted into the established formula, AUC were 0.885and 0.880 in validating groups respectively. Conclusion For parotid gland lesions with intratumorally cystic components, the characteristics of CT enhanced images can be used in differentiating benign/malignant lesions and adenolymphoma/pleomorphic adenoma. The predicting models basing on these features have good diagnostic effectiveness.

关 键 词：腮腺 多形性腺瘤 腺淋巴瘤 囊变 LOGISTIC回归分析 

分 类 号：R733.4[医药卫生—肿瘤]