摘要
舌诊作为一种独特的中医诊断手段,在疾病诊断、治疗及预防中发挥着不可替代的作用。针对传统舌诊主观性强和对医生经验依赖性高的问题,提出一种基于改进U-Net的舌象分割算法,通过引入条状池化和自适应空间金字塔池化(Atrous Spatial Pyramid Pooling,ASPP)对经典U-Net架构进行改进。首先对U-Net网络进行扩展,添加条状池化层以增强模型对舌象细节特征的捕捉能力,并引入ASPP模块以融合多尺度上下文信息。此外,通过数据集优化处理来增强模型的泛化能力。实验结果表明,改进后的U-Net模型在舌象分割任务上取得了显著的性能提升,提高了舌象分割的准确性,使得后续的舌象特征提取更加准确和一致,有助于标准化舌诊流程,具有较高的临床应用价值。
As a unique diagnostic method in traditional Chinese medicine,tongue diagnosis plays an irreplaceable role in the disease diagnosis,treatment,and prevention of diseases.To address the issues of subjectivity and heavy reliance on medical experience in traditional tongue diagnosis,we propose an improved U-Net-based tongue image segmentation algorithm.The classic U-Net architecture is enhanced by introducing strip pooling and Atrous Spatial Pyramid Pooling(ASPP).Specifically,the U-Net network is expanded with the addition of strip pooling layers to strengthen the model′s ability to capture detailed features of tongue images.Additionally,ASPP modules are incorporated to integrate multi-scale contextual information.Furthermore,the generalization capability of the model is enhanced through dataset optimization.Experimental results demonstrate that the improved U-Net model achieves significant performance improvements in tongue image segmentation tasks,increasing the accuracy of tongue image segmentation and making subsequent feature extraction more precise and consistent.This contributes to the standardization of the tongue diagnosis process and holds high clinical application value.
作者
唐满
魏兵
郭东恩
TANG Man;WEI Bing;GUO Dong-en(School of Computer and Software,Nanyang Institute of Technology,Nanyang 473000,China)
出处
《南阳理工学院学报》
2025年第4期8-15,共8页
Journal of Nanyang Institute of Technology
基金
河南省科技研发计划联合基金(235101610024)
南阳市科技攻关项目(24KJGG096)。
