皮肤镜影像结合特征分析算法权衡纵行黑甲活检必要性的临床研究  

A clinical study on the necessity of longitudinal melanonychia biopsy based on dermoscopy and feature analysis algorithm

作　　者：孙祺琳[1] 刘洋[1] 陈骏[1] 杨雅骊 李梦 徐慧[1] 刘科[1] SUN Qilin;LIU Yang;CHEN Jun;YANG Yali;LI Meng;XU Hui;LIU Ke(Department of Dermatology,Shanghai Ninth People's Hospital,Shanghai Jiao Tong University School of Medicine,Shanghai 200011,China)

机构地区：[1]上海交通大学医学院附属第九人民医院皮肤科,上海市200011

出　　处：《组织工程与重建外科》2021年第3期235-238,246,共5页Journal of Tissue Engineering and Reconstructive Surgery

基　　金：上海交通大学“科技创新专项资金”(YG2019QNA13);上海交通大学医学院附属第九人民医院基础研究助推计划(JYZZ035)。

摘　　要：目的基于纵行黑甲(Longitudinal melanonychia,LM)临床及皮肤镜图像,设计一种根据LM背景及条纹颜色鉴别甲下色素痣(Nail matrix nevus,NMN)和甲黑素活化(Melanocytic activation,MA)的分类算法,以指导临床是否需要进行甲活检。方法选取17例纵行黑甲患者的术前大体照片及皮肤镜图像,其中10例为甲黑素活化,7例为甲下色素痣。通过Lab颜色空间的单变量特征选择,分析正常及异常指(趾)甲区域内所计算特征的差异分布,对LM进行NMN和MA的二分类概率预测。结果通过算法对黑甲和正常区域的颜色均值方差进行分析显示,大体图的L亮度分量均值,大体图正常异常区域A颜色分量方差差值,皮肤镜图像正常异常区域L亮度分量均值差值,以及皮肤镜图像异常区域L亮度分量方差,对辅助诊断LM为MA或NMN有重要意义。结论通过特征提取、定量分析的黑甲算法模型,可有效鉴别MA和NMA,避免不必要的甲活检和甲损毁,有望成为一种术前检测和临床随访的无创、便捷方法。Objective Based on the dermoscope image and clinical photography of longitudinal melanonychia(LM),to design a new algorithm to differentiate the nail matrix nevus(NM)and melanocytic activation(MA)and furthermore indicate the necessity of nail biopsy according to the background and colour of LMs.Methods Preoperative gross photos and dermoscopic images of 17 LMs(10 MA and 7 NM)were chosen.Lab colour space univariate feature selection was done to analyze the differential distribution between the normal and abnormal district of an LM image and thus predict the dichotomous probability of NMN and MA.Results Algorithm for black and normal area by the color of the variance analysis,the gross figure L average luminance component and normal anomaly area A color difference of variance component,dermoscope image anomaly area of luminance component L mean difference and anomaly area skin brightness L weight variance of LM has important significance for the differentiation of MA and NMN.Conclusion The LM algorithm model based on feature extraction and quantitative analysis can effectively identify MA and NMN and avoid unnecessary nail biopsy and nail damage.It is expected to be a non-invasive and convenient method for preoperative detection and clinical follow-up.

关 键 词：纵行黑甲 甲下色素痣 甲黑素活化 皮肤镜图像 特征提取 

分 类 号：R758.72[医药卫生—皮肤病学与性病学]