基于卷积神经网络的臂丛神经超声图像分割方法  

Ultrasound image segmentation of brachial plexus via convolutional neural networks

作　　者：龙法宁 朱晓姝 甘井中[1] LONG Faning;ZHU Xiaoshu;GAN Jingzhong(School of Computer Science and Engineering,Yulin Normal University,Yulin 537000,China;Guangxi Universities Key Laboratory of Complex System Optimization and Big Data Processing,Yulin Normal University,Yulin 537000,China)

机构地区：[1]玉林师范学院计算机科学与工程学院,广西玉林537000 [2]玉林师范学院广西高校复杂系统优化与大数据处理重点实验室,广西玉林537000

出　　处：《合肥工业大学学报（自然科学版）》2018年第9期1191-1195,1296,共6页Journal of Hefei University of Technology：Natural Science

基　　金：广西壮族自治区自然科学基金资助项目(2013GXNSFAA019337);广西教育厅重点科研资助项目(2013ZD056);广西青年自然科学基金资助项目(2014GXNSFBA118010;2014GXNSFBA118268)

摘　　要：近几年卷积神经网络(convolutional neural network,CNN)在图像处理、语音识别、自然语言处理以及信息检索等领域得到广泛应用,颈部臂丛神经超声图像具有较低的信噪比、较低的对比度、模糊的边缘,其分割是一项富有挑战性的工作。文章针对目前臂丛神经超声图像手工标注的训练样本较少的情况,对U-Net模型进行改进,构建了一个适用于臂丛神经分割的卷积神经网络模型QU-Net,并选择BP图像数据库进行训练、测试。实验结果表明,与主流神经卷积网络分割算法SegNet、U-Net相比,QU-Net的图像分割结果具有更高的准确性。The convolutional neural network(CNN)has been widely used in the fields of speech recognition,picture processing,natural language processing and information retrieval in recent years.Because of low signal-to-noise ratio,low contrast and blurry boundaries,ultrasound image segmentation of brachial plexus is quite challenging.In this paper,a U-Net based CNN model named QU-Net model is established for the ultrasound image segmentation of brachial plexus,which is trained when the training data set is incomplete.Experimental results obtained via BP datasets show that this model performs better than several other algorithms such as SegNet and U-Net.

关 键 词：臂丛神经 图像分割 卷积神经网络(CNN) 深度学习 U-Net模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]