基于 U1-net 网络的放疗脑肿瘤靶区分割-学报杂志社

期刊信息

合肥工业大学（自然科学版），2023年8月，第46卷第8期：1070-1078

DOI: 10.3969/j.issn.1003-5060.2023.08.010

作者信息

张本健 $ ^{1} $，林辉 $ ^{1} $，郭栋 $ ^{2} $，王桂林 $ ^{1} $，胡敏 $ ^{2} $

(1. 合肥工业大学电子科学与应用物理学院，安徽合肥 230601；2. 合肥工业大学计算机与信息学院，安徽合肥 230601)

摘要和关键词

摘要: 文章基于全卷积神经网络（fully convolutional network，FCN）的 U-net 网络，并通过对 U-net 网络的调整，构建适用于脑肿瘤图像分割的 U1-net 网络。U1-net 网络由卷积层、最大池化层、反卷积层和激活函数 4 个部分组成。通过在公共数据集 BRATS 2015 上的实验验证了该网络的有效性。实验结果表明，该网络能适应脑肿瘤轮廓取得较好的分割效果，在脑肿瘤的完整肿瘤区、核心肿瘤区、增强肿瘤区的 Dice 相似系数（Dice similarity coefficient，DSC）分别为 0.95、0.85、0.83。

关键词: 深度学习(DL)；全卷积神经网络(FCN)；U1-net网络；BRATS 2015 数据集；脑肿瘤分割

Authors

ZHANG Benjian $ ^{1} $, LIN Hui $ ^{1} $, GUO Dong $ ^{2} $, WANG Guilin $ ^{1} $, HU Min $ ^{2} $

(1. School of Electronic Science and Applied Physics, Hefei University of Technology, Hefei 230601, China; 2. School of Computer Science and Information Engineering, Hefei University of Technology, Hefei 230601, China)

Abstract and Keywords

Abstract: This paper is based on the U-net network of the fully convolutional network (FCN), and by adjusting the U-net network, a U1-net network suitable for brain tumor image segmentation is constructed. The U1-net network is composed of four parts: convolution layer, max pooling layer, deconvolution layer and activation function. Through the experimental verification on the public data set BRATS 2015, the validity of the model is verified. The experimental results show that the model can adapt to the contour of brain tumors and achieve a good segmentation effect. In addition, the Dice similarity coefficient (DSC) values of 0.95, 0.85 and 0.83 are obtained in the complete tumor region, core tumor region and enhanced tumor region of brain tumors, respectively.

Keywords: deep learning (DL); fully convolutional network (FCN); U1-net network; BRATS 2015 data set; brain tumor segmentation

基金信息

国家自然科学基金资助项目（61672202；U1613217）

合肥工业大学学报(自然科学版)

导航菜单

基于 U1-net 网络的放疗脑肿瘤靶区分割