基于高频超声的局限性硬皮病皮损分期预测模型的建立及验证  

Construction and validation of a prediction model for staging of localized scleroderma lesions based on high-frequency ultrasound

作　　者：柴可 喻江帆 林彩虹 唐冰思 尤睿璇 曾茁桐 史雅倩 邱湘宁[1] 湛意[1] 张桂英[1] 刘明辉[3] 肖嵘[1] Chai Ke;Yu Jiangfan;Lin Caihong;Tang Bingsi;You Ruixuan;Zeng Zhuotong;Shi Yaqian;Qiu Xiangning;Zhan Yi;Zhang Guiying;Liu Minghui;Xiao Rong(Department of Dermatology and Venereology,The Second Xiangya Hospital of Central South University,Institute of Dermatology and Venereology of Central South University,Hunan Clinical Medicine Research Center for Major Skin Diseases and Skin Health,Changsha 410011,China;Department of Diagnostic Ultrasound,Pingxiang No.2 People′s Hospital,Pingxiang 337000,Jiangxi,China;Department of Diagnostic Ultrasound,The Second Xiangya Hospital of Central South University,Changsha 410011,China)

机构地区：[1]中南大学湘雅二医院皮肤性病科、中南大学皮肤性病研究所、皮肤重大疾病与皮肤健康湖南省临床医学研究中心,长沙410011 [2]萍乡市第二人民医院超声诊断科,萍乡337000 [3]中南大学湘雅二医院超声诊断科,长沙410011

出　　处：《中华皮肤科杂志》2023年第11期1008-1015,共8页Chinese Journal of Dermatology

基　　金：国家自然科学基金(82373486、82073449、82003363、82203932)。

摘　　要：目的分析局限性硬皮病患者的临床特征与皮损高频超声特点,构建使用皮损高频超声特征无创性预测皮损分期的模型并进行验证。方法回顾性纳入2021年2月1日至2023年2月28日期间于中南大学湘雅二医院皮肤科就诊的局限性硬皮病患者,收集患者的临床资料、共85处皮损的高频超声及病理数据,按照患者入组的时间顺序分为训练队列和验证队列。使用单因素分析和多因素logistic回归筛选训练队列皮损分期的差异因素,构建回归方程,建立Nomogram预测模型。通过Bootstrap法对模型进行内部验证,采用校准曲线及接收者操作特征(ROC)曲线进一步评估模型在训练队列和验证队列中的预测效能。结果训练队列纳入60例局限性硬皮病患者,男性16例,女性44例,年龄[M(Q1,Q3),22.0(10.0,39.2)]岁,包括28例炎症水肿期皮损与32例硬化萎缩期皮损;验证队列纳入25例患者,男性8例,女性17例,年龄18.0(7.0,30.0)岁,包括9例炎症水肿期皮损和16例硬化萎缩期皮损。单因素分析结果提示,训练队列炎症水肿期与硬化萎缩期组患者年龄、性别、皮损部位的分布差异均无统计学意义(均P>0.05);硬化萎缩期皮损组与炎症水肿期皮损组相比,发病时长≥2年的患者比例更高(20/32例比10/28例,χ^(2)=4.29,P=0.038)、皮损皮下脂肪层厚度更薄[1.4(0.0,26.0)mm比1.8(0.1,14.3)mm,Z=-2.14,P=0.032]、皮损脂肪层厚度较对照脂肪层减少值更大[1.8(0.5,11.0)mm比0.3(-1.9,8.0)mm,P<0.001]、皮损与对照皮肤弹性值之比更高[2.9(1.8,6.9)比1.8(1.1,5.9),Z=-4.34,P<0.001]、皮损活动超声评分更高[5.0(3.0,8.0)比3.0(0.0,5.0),Z=-4.76,P<0.001]。多因素logistic逐步回归结果显示发病时长≥2年(P=0.032)、皮损与对照皮肤弹性值之比(P=0.019)、皮损活动超声评分(P=0.004)和皮损脂肪层厚度较对照脂肪层减少值(P=0.013)的增高对定义硬化萎缩期皮损有正向影响。根据回归分析结果,Nomogram模型纳入�Objective To analyze clinical characteristics and high-frequency ultrasound features of localized scleroderma,and to construct and validate a non-invasive prediction model for staging of skin lesions based on the high-frequency ultrasound features.Methods Patients with localized scleroderma were retrospectively collected from the Department of Dermatology and Venereology,Second Xiangya Hospital of Central South University from February 1,2021 to February 28,2023,and clinical data as well as high-frequency ultrasound and pathologic features of 85 lesions from these patients were analyzed.Lesions were divided into modeling cohort and validation cohort according to the chronological order of patient enrollment.The univariate analysis and multivariable logistic regression models were used to analyze the independent influential factors in the staging of localized scleroderma lesions in the modeling cohort,construct the regression equation,and to build a nomogram prediction model.The Bootstrap validation method was used for internal validation,and the predictive performance of the nomogram model in the modeling cohort and validation cohort was further evaluated by the calibration curve and receiver operating characteristic(ROC)curve.Results In the modeling cohort,60 patients with localized scleroderma,including 16 males and 44 females,were enrolled,with the age[M(Q1,Q3)]being 22.0(10.0,39.2)years,and there were 28 lesions in the oedematous phase and 32 lesions in the fibrotic and atrophic phase;in the validation cohort,25 patients with localized scleroderma,including 8 males and 17 females,were enrolled,with the age being 18.0(7.0,30.0)years,and there were 9 lesions in the oedematous phase and 16 lesions in the fibrotic and atrophic phase.Univariate analysis in the modeling cohort showed no significant differences in the age and gender of patients or the location of lesions between the oedematous phase group and the fibrotic and atrophic phase group(all P>0.05);compared with the oedematous phase group,the fibrotic and

关 键 词：硬皮病 局部性 超声检查 预测 高频超声 皮损分期 Nomogram预测模型 

分 类 号：R593.25[医药卫生—内科学]