按照数据集进行划分:
关于自动分割工具,婴儿脑MR图像来自单个时间点,其中纵向数据集不可用,因此必须开发不针对纵向数据集的分割工具,目前提出了一些机器学习方法,但这些方法的效果并不令人满意。
iseg-2017:
MICCAI iseg-2017挑战赛结果: 第一名为MSL_SKKU
前五名的结果为:
与数据集相关的论文:
Bui T D, Shin J, Moon T. 3d densely convolution networks for volumetric segmentation[J]. arXiv preprint arXiv:1709.03199, 2017.
---10引用 采用DenseNet 挑战赛中排名第一 队伍名:MSL-SKKU
取小数点后3位的情况下9项指标6项第一
Dolz J, Desrosiers C, Wang L, et al. Deep CNN ensembles and suggestive annotations for infant brain MRI segmentation[J]. arXiv preprint arXiv:1712.05319, 2017.
---6引用 采用SemiDenseNet 在某些指标中排名第一或者第二,结果如下:
Dolz J, Ayed I B, Yuan J, et al. Isointense infant brain segmentation with a hyper-dense connected convolutional neural network[C]//Biomedical Imaging (ISBI 2018), 2018 IEEE 15th International Symposium on. IEEE, 2018: 616-620.
---3引用 采用DenseNet 9个指标中6个排名前三
该文章基于另外两篇文章:
Konstantinos Kamnitsas et al., "Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation", Medical image analysis, vol. 36, pp. 61-78, 2017.
Jose Dolz et al., "3D fully convolutional networks for subcortical segmentation in MRI: A large-scale study", NeuroImage, 2017.
Dolz J, Gopinath K, Yuan J, et al. HyperDense-Net: A hyper-densely connected CNN for multi-modal image segmentation[J]. arXiv preprint arXiv:1804.02967, 2018.
取小数点后两位的情况下,都为9项指标7项第一
对于MRBrainS2013:
提交时间为18.02.16,提交时为第一名,目前排名第6,对比结果如下:
| HyperDenseNet(top-6) | 0.8633 | 1.34 | 6.19 | 0.8946 | 1.78 | 6.03 | 0.8342 | 2.26 | 7.31 |
| XMU_SmartDSP2(top-1) | 0.865 | 1.29 | 5.75 | 0.899 | 1.73 | 5.47 | 84.8 | 1.84 | 6.83 |
Dolz J, Ayed I B, Yuan J, et al. HyperDense-Net: A hyper-densely connected CNN for multi-modal image semantic segmentation[J]. arXiv preprint arXiv:1710.05956, 2017.
---未了解(上一篇的效果较好)
Fonov V, Doyle A, Evans A C, et al. NeuroMTL iSEG challenge methods[J]. bioRxiv, 2018: 278465.
---排名较靠前 队伍名:NeuroMTL
Sanroma G, Benkarim O M, Piella G, et al. Learning to combine complementary segmentation methods for fetal and 6-month infant brain MRI segmentation[J]. Computerized Medical Imaging and Graphics, 2018, 69: 52-59.
---非顶会,效果不好但拿不到数据集
Zeng G, Zheng G. Multi-stream 3D FCN with multi-scale deep supervision for multi-modality isointense infant brain MR image segmentation[C]//Biomedical Imaging (ISBI 2018), 2018 IEEE 15th International Symposium on. IEEE, 2018: 136-140.
---排名第三 队伍名:Bern_IPMI 采用FCN
Wang Z, Zou N, Shen D, et al. Global Deep Learning Methods for Multimodality Isointense Infant Brain Image Segmentation[J]. arXiv preprint arXiv:1812.04103, 2018.
-----预印本,目前最新效果最好的的论文(WM与GM为最佳性能,CSF具有可比性),提交时间为2018.12.10,未参与挑战赛排名,实验结果为:
| WM | GM | CSF | |
| top-1 | 0.901 | 0.919 | 0.958 |
| 论文 | 0.9044 | 0.9219 | 0.9557 |
使用的数据集为 iseg-2017,使用的网络为FCNN,实验结果接近第一名,实验结果如下:
数据集来源于BCP项目,该数据集作为iseg-2017挑战赛数据集的一部分(总数,训练测试集数一样),结果比挑战赛TOP-1低2%,但参数减半。
推荐的文章为:但3篇文章都为预印本
1.Bui T D, Shin J, Moon T. 3d densely convolution networks for volumetric segmentation[J]. arXiv preprint arXiv:1709.03199, 2017.
---10引用 采用DenseNet 挑战赛中排名第一 队伍名:MSL-SKKU
取小数点后3位的情况下9项指标6项第一
2.Dolz J, Gopinath K, Yuan J, et al. HyperDense-Net: A hyper-densely connected CNN for multi-modal image segmentation[J]. arXiv preprint arXiv:1804.02967, 2018.
--取小数点后2为的情况下都为9项指标7项第一
3.Wang Z, Zou N, Shen D, et al. Global Deep Learning Methods for Multimodality Isointense Infant Brain Image Segmentation[J]. arXiv preprint arXiv:1812.04103, 2018.
-----预印本,目前最新效果最好的的论文(WM与GM为最佳性能,CSF具有可比性),提交时间为2018.12.10,未参与挑战赛排名
NDAR数据集
Wang L, Li G, Shi F, et al. Volume-based analysis of 6-month-old infant brain MRI for autism biomarker identification and early diagnosis[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, Cham, 2018: 411-419.
采用U-Net,数据集来源于NDAR,包含18名受试者,分割结果很好------未找到数据集
neobrains(类别较多)
第一名排名结果如下(在大多数指标中排名第一)
其他
MICCAI挑战赛底部文章:
[1]. Li Wang, Yaozong Gao, Feng Shi, Gang Li, John H. Gilmore, Weili Lin, Dinggang Shen. LINKS: Learning-based multi-source IntegratioN frameworK for Segmentation of infant brain images, Neuroimage, 108, 160-172, 2015.
[2]. Li Wang, Feng Shi, Yaozong Gao, Gang Li, John H. Gilmore, Weili Lin, Dinggang Shen. Integration of Sparse Multi-modality Representation and Anatomical Constraint for Isointense Infant Brain MR Image Segmentation, Neuroimage, 89, 152-164, 2014.
[3]. Wenlu Zhang, Rongjian Li, Houtao Deng, Li Wang, Weili Lin, Shuiwang Ji, Dinggang Shen. Deep Convolutional Neural Networks for Multi-Modality Isointense Infant Brain Image Segmentation, Neuroimage, 108, 214–224, 2015.
综述文章:
Li G, Wang L, Yap P T, et al. Computational neuroanatomy of baby brains: A review[J]. Neuroimage, 2018.
---引用9 综述文章