传统方法和深度学习用于不同模态心脏医学图像的分割研究进展
常博, 孙灏芸, 高清宇, 王丽嘉

Research Progress on Cardiac Segmentation in Different Modal Medical Images by Traditional Methods and Deep Learning
CHANG Bo, SUN Haoyun, GAO Qingyu, WANG Lijia
表3 基于心脏CT图像的分割网络总结
Table 3 Summary of different segmentation networks based on cardiac CT images
方法 时间 学习框架 数据集 Dice系数 Hausdorff距离(HD)/mm
左心室
(LV)
右心室
(RV)
心肌
(Myo)
左心室
(LV)
右心室
(RV)
心肌
(Myo)
CNN[93] 2016 - 60例临床病例 0.85 - - - - -
Combining faster R-CNN and
U-net[54]
2018 PyTorch MM-WHS2017 0.879 0.902 0.822 - - -
CNN[102] 2018 TensorFlow 11例临床病例 0.878 0.829 - - - -
Hybrid loss guided CNN[65] 2018 TensorFlow MM-WHS2017 0.8680 0.7143 0.665 - - -
CNN and anatomical label
configurations[94]
2018 Caffe MM-WHS2017 0.918 0.909 0.881 - - -
3D deeply-supervised U-Net[55] 2018 - MM-WHS2017 0.893 0.810 0.837 - - -
DL and shape context[59] 2018 Keras MM-WHS2017 0.935 0.825 0.879 - - -
Multi-planar deep segmentation
networks[99]
2018 TensorFlow MM-WHS2017 0.904 0.883 0.851 - - -
3D CNN[103] 2018 TensorFlow MM-WHS2017 0.923 0.857 0.856 - - -
Two-stage 3D U-net[56] 2018 TensorFlow MM-WHS2017 0.800 0.786 0.729 - - -
Multi-depth fusion network[58] 2019 TensorFlow MICCAI 2017全心
CT数据集
0.944 0.895 0.889 - - -
3D deeply supervised attention
U-net[57]
2020 MATLAB 100例临床病例 0.916 - - 6.840 - -
DL[66] 2020 - 1100例临床数据 - - 0.883 - - 13.4
Unet-GAN[98] 2021 PyTorch MM-WHS2017 整体平均0.889
Multiple GAN guided by
Self-attention mechanism[97]
2021 - MM-WHS2017 0.814 - 0.669 - - -
AttU_Net_conv1_5Mffp[62] 2021 PyTorch MM-WHS2017 0.907 0.842 0.906 - - -
PC-Unet[60] 2021 - 20例临床数据 0.885 - - 7.05 - -
Computer graphics imaging
and DL[129]
2022 - 130例临床数据 - 0.81~0.95 - - - -
DRLSE[25] 2022 - 5例临床数据 0.9253 - - 7.874 - -
4D contrast-enhanced[104] 2022 PyTorch 1509例临床数据 整体平均0.8 - - -
MRDFF[95] 2022 - MM-WHS2017 0.899 0.823 - - - -
Transnunet[64] 2022 - MM-WHS2017 0.921 - - - - -
Self-attention mechanism[45] 2023 TensorFlow 96例临床病例 - - 0.9202 - - -