摘要
甲病的诊断往往依赖经验丰富的临床专家,而近年来在皮肤病诊断领域中取得显著进展的人工智能(AI)技术,在甲病的识别和辅助诊断方面同样前景广阔。本文总结了甲病的图像特征(如颜色与形态变化)及其对图像处理算法提出的要求,并综述了常见甲病中AI模型的应用实例。针对甲真菌病,多项研究已开发多种基于不同图像类型的AI诊断模型;在甲银屑病中应用的AI模型则主要关注于甲银屑病严重程度指数评分自动化;而针对甲黑线疾病的相关AI模型则通过分割和(或)分类的方式实现疾病识别。此外,除了甲疾病,AI模型还可通过分析甲图像中的微血管和血红蛋白变化,辅助诊断糖尿病等全身性疾病。尽管现有模型各有局限,但随着高质量数据集的积累、算法的不断优化和临床应用规范的完善,AI有望成为甲病辅助诊断中不可或缺的工具。
Diagnosis of nail disorders often relies on the expertise of experienced clinicians.However,artificial intelligence(AI),which has achieved significant progress in the field of diagnosis of skin diseases in recent years,also holds great promise for the identification and auxiliary diagnosis of nail diseases.This review summarizes the imaging features of nail disorders,such as changes in color and morphology,and the corresponding requirements they pose for image processing algorithms.This paper also reviews examples of AI model applications in the auxiliary diagnosis of common nail diseases.For onychomycosis,multiple studies have developed various AI models based on different image types.In the case of nail psoriasis,existing AI models have mainly focused on automating severity scoring systems.For melanonychia,relevant AI models typically identify disease patterns through segmentation and/or classification approaches.In addition to nail diseases,AI models can also assist in diagnosing systemic conditions such as diabetes by analyzing microvascular changes and hemoglobin distribution in nail images.Although current models have their limitations,the continued accumulation of high-quality datasets,ongoing algorithmic advancements,and the development of standardized clinical applications are expected to make AI an indispensable tool in the auxiliary diagnosis of nail diseases.
作者
何蒙文
麦思恩
马寒
HE Mengwen;MAI Sien;MA Han(Fifth Affiliated Hospital of Sun Yat-sen University,Zhuhai 519000,China)
出处
《皮肤性病诊疗学杂志》
2025年第5期363-370,共8页
Journal of Diagnosis and Therapy on Dermato-venereology
基金
国家重点研发计划(2024YFB3614303)
中山大学附属第五医院院内IIT资助(YNZZ2022-02)。
关键词
甲病
人工智能
辅助诊断
卷积神经网络
nail diseases
artificial intelligence
auxiliary diagnosis
convolutional neural network
