基于超声图像特征的Nomogram模型可准确定性诊断颈部肿大淋巴结  

Nomogram model based on ultrasound image features for accurate qualitative diagnosis of enlarged cervical lymph nodes

作　　者：袁虹 文文 阮娟 赵海娜[1] 刘晶焰 彭玉兰[1] YUAN Hong;WEN Wen;RUAN Juan;ZHAO Haina;LIU Jingyan;PENG Yulan(Department of Ultrasound,West China Hospital,Sichuan University,Chengdu 610041,China)

机构地区：[1]四川大学华西医院超声科,四川成都610041

出　　处：《分子影像学杂志》2025年第4期405-411,共7页Journal of Molecular Imaging

基　　金：四川省科技计划项目(2023NSFSC1722)。

摘　　要：目的建立基于超声图像特征的颈部淋巴结(CLN)定性诊断模型。方法回顾性选择2020年1月~2023年12月在四川大学华西医院超声科行淋巴结穿刺的患者2697例,共有颈部肿大淋巴结3014例,以病理学结果为金标准,分为良性病理1489例,恶性病理1525例。收集患者一般资料、二维灰阶及彩色多普勒超声图像,由2位超声医师读图记录淋巴结图像特征。将患者以7∶3的比例随机分割为训练队列及验证队列。在验证组队列中,通过单因素分析筛选CLN良恶性相关特征,采用逻辑回归进行多因素分析确定候选变量,构建Nomogram模型。分别在训练集及测试集绘制ROC曲线,计算敏感度、特异度、准确度评估模型的诊断效能,绘制校准曲线评估模型预测值和实际观测值之间的差异,绘制临床决策曲线分析评价其临床有效性。结果共有10个变量纳入Nomogram模型,模型训练组的曲线下面积(AUC)为0.942,验证组的AUC为0.925,常规超声定性诊断淋巴结的AUC为0.656,差异有统计学意义(P<0.001)。淋巴结长径、短径、长短径比、皮质回声模式、皮质是否均质、皮髓质分界、淋巴门显示、边缘、形状、钙化是CLN定性诊断的重要超声特征。结论基于超声图像的Nomogram模型能够提供CLN定性诊断,较常规超声诊断具有明显优越性,为临床CLN肿大患者下一步诊疗方案提供重要参考。Objective To establish a qualitative diagnostic nomogram for cervical lymph nodes(CLNs)based on ultrasonographic characteristics.Methods A retrospective study was conducted on 2,697 patients who underwent fine needle aspiration of CLNs at West China Hospital of Sichuan University from January 2020 to December 2023.The analysis encompassed a total of 3014 enlarged CLNs,which were categorised into 1489 benign and 1525 malignant cases based on pathological findings.Clinicopathplogical information,B-mode ultrasound images,and color Doppler ultrasound images were systematically collected.Two experienced radiologists independently reviewed the imaging features of the lymph nodes.The patients were randomly assigned to a training cohort and a validation cohort at a ratio of 7:3.In the validation cohort,univariate analysis and multivariate logistic regression analysis were performed to identify candidate features associated with benign or malignant CLN.The construction of the nomogram model was facilitated by the identified candidate features.ROC curves were plotted in both the training and validation sets to evaluate diagnostic performance by calculating sensitivity,specificity,and accuracy.Calibration curves were utilised to evaluate the concordance between predicted probabilities and observed outcomes,while clinical decision curve analysis was employed to assess the clinical utility of the model.Results Based on univariate and multivariate analysis,ten features were integrated into the nomogram model ultimately.This model exhibited an AUC of 0.942 in the training cohort and 0.925 in the validation cohort,significantly exceeding the performance of conventional ultrasound qualitative diagnosis(AUC=0.656,P<0.001).The ultrasound features incorporated into the model included the long-axis diameter,short-axis diameter,long-to-short axis ratio,cortical echogenicity pattern,cortical homogeneity,corticomedullary demarcation,hilum visualization,margin,shape,and calcification.Conclusion The ultrasonography-based nomogram demonstr

关 键 词：颈部淋巴结 超声 NOMOGRAM 

分 类 号：R445.1[医药卫生—影像医学与核医学]