基于Swin Transformer的皮肤病诊断研究  

Research on Dermatological Disease Diagnosis Based on Swin Transformer

作　　者：申雨婧 张仲荣 宋平安 

机构地区：[1]兰州交通大学数理学院,甘肃 兰州 [2]兰州石化总医院胸外科,甘肃 兰州

出　　处：《应用数学进展》2025年第2期34-39,共6页Advances in Applied Mathematics

基　　金：本研究受甘肃省科技计划项目(24YFFA055, 22JR5RA797)、甘肃省云计算重点实验室开放课题(2023KFKT-005)与甘肃省重点人才项目(“东数西算”场景下的后量子数据加密传输机制研究)的资助。

摘　　要：皮肤病是世界范围内最常见的疾病之一。准确、及时、有效的皮肤病图像分类研究对皮肤病诊断具有重要意义,高精度分类算法研究是该领域的热点和难点。近年来,深度学习算法在皮肤病诊断领域展现出了巨大的发展潜力,具有广阔的应用前景。文章采用基于Transformer模型改进的Swin Transformer模型,构建了基于Swin Transformer的皮肤病诊断模型,并将该模型在HAM10000数据集上进行实验验证。研究证明,该模型显著提高了皮肤病诊断的准确性,有望推动皮肤病诊断领域朝着更加高效、精准的方向迈进,为皮肤病诊断领域的临床实践提供了更有力的实验支撑。Dermatological diseases are among the most common diseases worldwide. Accurate, timely, and effective classification of dermatological disease images is of great significance for the diagnosis of dermatological diseases, and the research on high-precision classification algorithms is a hot and difficult issue in this field. In recent years, deep learning algorithms have shown great potential for development in the field of dermatological disease diagnosis, with broad application prospects. This paper adopts the Swin Transformer model, an improvement based on the Transformer model, to construct a dermatological disease diagnosis model based on the Swin Transformer. The model is experimentally verified on the HAM10000 dataset. The study proves that the model significantly improves the accuracy of dermatological disease diagnosis and is expected to promote the field of dermatological disease diagnosis in a more efficient and precise direction, providing stronger experimental support for clinical practice in this field.

关 键 词：Swin Transformer 自注意力机制 皮肤病诊断 

分 类 号：R75[医药卫生—皮肤病学与性病学]