摘要
传统皮肤镜检查高度依赖医生的临床经验,并且皮肤镜图像本身的复杂性给临床诊断提出了巨大的挑战。针对以上情况,研发出一款带有人工智能辅助诊断功能的偏振皮肤镜系统。该系统使用偏振照明成像技术,经过目镜系统放大,由可连续工作的工业相机将大量实时抓拍的皮肤镜裸数据图像传输至上位机。通过上位机内置的图像处理模块将图像进行预处理,然后通过训练好的卷积神经网络VGG16将皮肤镜图像分类。最后,用户可以通过皮肤镜诊断平台查看诊断结果。此系统可以更加清晰有效地观察深层皮肤组织病变组织形态,由系统采集的皮肤镜图像的分类准确性达到93.6%,可用于临床诊疗。
Traditional skin microscopy heavily relies on the clinical experience of physicians,and the complexity of skin microscopy images poses significant challenges to clinical diagnosis.In response to this situation,a polarized skin microscopy system with artificial intelligence-assisted diagnostic functionality has been developed.This system utilizes polarized illumination imaging technology.After magnification through an eyepiece system,a continuously operating industrial camera captures a large volume of real-time raw skin microscopy data images and transfers them to the host computer.The images are preprocessed using the built-in image processing module in the host computer and are then classified using a trained convolutional neural network,VGG16,f or skin microscopy images.Finally,users can view the diagnostic results through the skin microscopy diagnostic platform.This system allows for a clearer and more effective observation of the morphological characteristics of lesions in deep skin tissues.The classification accuracy of skin microscopy images captured by the system reaches 93.6%,making it suitable for clinical diagnosis and treatment.
作者
武正国
庞春颖
张茜然
孙嘉灵
WU Zhengguo;PANG Chunying;ZHANG Xiran;SUN Jialing(School of Life Science and Technology,Changchun University of Science and Technology,Changchun 130022)
出处
《长春理工大学学报(自然科学版)》
2024年第2期114-120,共7页
Journal of Changchun University of Science and Technology(Natural Science Edition)
基金
吉林省科技厅项目(20220204127YY,20210401160YY)。
关键词
皮肤镜
偏振光
皮肤镜图像
人工智能辅助诊断
dermatoscopy
polarized light
dermatoscopy images
artificial intelligence-assisted diagnosis