光学学报, 2020, 40 (6): 0610001, 网络出版: 2020-03-06
多模态磁共振脑肿瘤图像自动分割算法研究 下载: 1791次
Automatic Segmentation Algorithm for Multimodal Magnetic Resonance-Based Brain Tumor Images
图 & 表
图 2. 3D卷积与输入体素块的点乘运算示意图。(a)普通3D卷积运算;(b)膨胀率r=2的3D膨胀卷积运算
Fig. 2. Dot product operation diagram of 3D convolution with input voxel block. (a) Original 3D convolution operation; (b) 2-dilated 3D convolution operation
图 4. 收缩路径中输入图片降采样与混合膨胀卷积模块的感受野示意图
Fig. 4. Sketch map of input images sub-sampling and HDC receptive fieldNote:In layer 3 4 5, the size of images is enlarged to scale for better demonstration.
图 8. 算法流程图与各阶段图像处理结果
Fig. 8. Algorithm flowchart and image processing results of each stage
表 1BraTS 2017数据集中各类别体素占总体素的比值
Table1. Ratio of different types of voxels to total voxel in the BraTS 2017 dataset
|
表 28种模型的评价指标数据对比
Table2. Comparison of evaluation index data of eight models
|
何承恩, 徐慧君, 王忠, 马丽萍. 多模态磁共振脑肿瘤图像自动分割算法研究[J]. 光学学报, 2020, 40(6): 0610001. Cheng'en He, Huijun Xu, Zhong Wang, Liping Ma. Automatic Segmentation Algorithm for Multimodal Magnetic Resonance-Based Brain Tumor Images[J]. Acta Optica Sinica, 2020, 40(6): 0610001.