摘要
皮肤镜是一种观察活体皮肤表面以下微细结构和色素的无创性显微图像分析技术,已经成为皮肤癌以及各种皮肤疾病的筛选和诊断的有效工具。本文针对黄色人种皮肤镜图像,提出了一套带有质量评价功能的皮肤镜图像自动分析系统。该系统首先采用预处理方法对皮肤镜图像进行质量提升;然后采用基于多模式分类的方法实现皮肤镜图像的自适应分类;最后采用组合神经网络模型,实现皮损的分类识别。对180幅黄色人种皮肤镜图像进行分类结果统计,获得了87.8%的敏感度和95.8%的特异度,较其他方法具有更好的分类准确率。
Dermoscopy allows a better visualization of the skin surface and subsurface structures,and is beneficial to diagnose many skin diseases in clinical applications. In this paper,a computer aided diagnosis system with image quality assessment is proposed for dermoscopy images of the Xanthous race. First-ly,the dermorcopy image quality is assessed and improved through preprocessing methods. Secondly,the dermoscopy image is adaptively segmented based on multi- pattern classification. Lastly,the lesion is classified using an neural network ensemble model. A series of experiments are carried out on the Xanthous race dataset with 180 dermoscopy images. The results show that our method with gains sensitivity of87. 8% and the specificity of 95. 8% which is superior to other compared methods in comparison.
作者
谢凤英
姜志国
孟如松
XIE Fengying JIANG Zhiguo MENG Rusong(Image Processing Center, School of Astronautics, Beihang University, Beijing 100191, China General Hospital of the Air Force, PLA, Beijing 100142, China)
出处
《中国体视学与图像分析》
2016年第3期253-262,共10页
Chinese Journal of Stereology and Image Analysis
基金
国家自然科学基金(61471016
61371134
61271436
61027004
61071138)
关键词
皮肤镜图像
计算机辅助诊断
图像质量评价
图像分割
皮损分类
dermoscopy image
computer-aided diagnosis
image quality assessment
image segmentation
lesion classification