基于2.5D网络和尺度注意力感知的肝脏与肿瘤分割
A 2.5D Network and Scale Attention Awareness for Liver and Tumor Segmentation
摘要: 深度学习技术已被广泛应用在肝脏与肿瘤的分割任务中。但是，现有的3D网络模型大都忽略了医学图像横断面的像素距离过大的问题，直接使用3D卷积操作难以学习到准确的三维空间信息。此外，肿瘤形状大小高度可变的特点使得分割肿瘤更具挑战性。针对第一个问题，本文提出改进后的2.5D ResNet34对肝脏和肿瘤特征进行编码，提高模型对三维空间信息的建模能力。同时，利用DSC损失函数来提高模型对整体结构的分割能力。针对第二个问题，本文提出尺度注意力感知模块，通过建模不同尺度特征下的局部和全局三维空间信息，以有效地整合低级上下文信息和高级区域语义信息，从而实现精准的肝脏和肿瘤分割。本文所提出的方法在Liver Tumor Segmentation (LiTS)数据集上测试了肝脏与肿瘤的分割性能，其中肝脏分割的DSC为96.4%，肿瘤分割的DSC为72.3%，并与近三年的模型相比，本文提出的方法在肝脏和肿瘤分割中表现最好。
Abstract: Deep learning technology has been widely used in the segmentation task of liver and tumor. However, the existing 3D network models mostly ignore the problem of large pixel distances across medical images, and it is difficult to learn accurate three-dimensional spatial information directly using 3D convolution operations. In addition, the highly variable shape and size of tumors make segmenting tumors more challenging. Aiming at the first problem, this paper proposes an improved 2.5D ResNet34 to encode liver and tumor features, and improve the modeling ability to model three-dimensional spatial information. At the same time, the DSC loss function is utilized to improve the modeling ability to segment the overall structure. In response to the second problem, this paper proposes a scale attention awareness module, which can effectively integrate low-level context information and high-level regional semantic information by modeling local and global three-dimensional spatial information under different scale features, thereby achieving accurate segmentation of liver and tumor. The proposed method tested the performance on the Liver Tumor Segmentation (LiTS) dataset. The DSC of liver segmentation was 96.4%, and the DSC of tumor segmentation was 72.3%. Moreover, compared with the methods in the past three years, the proposed method performs best in liver and tumor segmentation.
文章引用：李家健, 黄国恒. 基于2.5D网络和尺度注意力感知的肝脏与肿瘤分割[J]. 计算机科学与应用, 2022, 12(1): 199-210. https://doi.org/10.12677/CSA.2022.121021

参考文献

[1] Ferlay, J., Shin, H.R., Bray, F., Forman, D., Mathers, C. and Parkin, D.M. (2010) Estimates of Worldwide Burden of Cancer in 2008: GLOBOCAN 2008. International Journal of Cancer, 127, 2893-2917.
[2] Müller, S.A., Maier-Hein, L., Mehrabi, A., Pianka, F., Rietdorf, U., Wolf, I. and Schmied, B.M. (2007) Creation and Establishment of a Respiratory Liver Motion Simulator for Liver Interventions. Medical Physics, 34, 4605-4608.
[3] 马金林, 邓媛媛, 马自萍. 肝脏肿瘤CT图像深度学习分割方法综述[J]. 中国图象图形学报, 2020, 25(10): 2024-2046.
[4] Dou, Q., Chen, H., Jin, Y.M., Yu, L.Q., Qin, J. and Heng, P.A. (2016) 3D Deeply Supervised Network for Automatic Liver Segmentation from CT Volumes. International Conference on Medical Image Computing and Computer-Assisted Intervention, Athens, 17-21 October 2016, 149-157. [Google Scholar] [CrossRef] 
[5] Bi, L., Kim, J., Kumar, A. and Feng, D. (2017) Automatic Liver Lesion Detection Using Cascaded Deep Residual Networks.
https://arxiv.org/abs/1704.02703
[6] Long, J., Shelhamer, E. and Darrell, T. (2015) Fully Convolutional Net-works for Semantic Segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39, 640-651. [Google Scholar] [CrossRef] 
[7] Ronneberger, O., Fischer, P. and Brox, T. (2015) U-Net: Convolutional Networks for Biomedical Image Segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention, Munich, 5-9 October 2015, 234-241. [Google Scholar] [CrossRef] 
[8] Milletari, F., Navab, N. and Ahmadi, S.A. (2016) V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation. 2016 Fourth International Con-ference on 3D Vision (3DV), Stanford, 25-28 October 2016, 565-571. [Google Scholar] [CrossRef] 
[9] He, K.M., Zhang, X.Y., Ren, S.Q. and Sun, J. (2016) Deep Residual Learning for Image Recognition. 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, 27-30 June 2016, 770-778. [Google Scholar] [CrossRef] 
[10] Li, X., Hao, C., Qi, X., Qi, D., Fu, C.W. and Pheng-Ann, H. (2017) H-DenseUNet: Hybrid Densely Connected UNet for Liver and Liver Tumor Segmentation from CT Volumes. IEEE Transactions on Medical Imaging, 37, 2663-2674.
[11] Han, X. (2017) Automatic Liver Lesion Segmentation Using a Deep Convolutional Neural Network Method.
https://arxiv.org/abs/1704.07239
[12] Jin, Q., Meng, Z., Sun, C., Wei, L. and Su, R. (2018) RA-UNet: A Hybrid Deep Attention-Aware Network to Extract Liver and Tumor in CT Scans.
https://arxiv.org/abs/1811.01328
[13] Chen, L.C., Zhu, Y., Papandreou, G., Schroff, F. and Adam, H. (2018) Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation. European Conference on Computer Vision, Munich, 8-14 September 2018, 833-851. [Google Scholar] [CrossRef] 
[14] Chen, L.C., Papandreou, G., Schroff, F. and Adam, H. (2017) Rethinking Atrous Convolution for Semantic Image Segmentation.
https://arxiv.org/abs/1706.05587
[15] Tang, Y.B., Tang, Y.X., Zhu, Y.Y., Xiao, J. and Summers, R.M. (2020) E2Net: An Edge Enhanced Network for Accurate Liver and Tumor Segmentation on CT Scans. International Conference on Medical Image Computing and Comput-er-Assisted Intervention, Lima, 4-8 October 2020, 512-522. [Google Scholar] [CrossRef] 
[16] Gao, S., Cheng, M.M., Zhao, K., Zhang, X.Y. and Torr, P. (2019) Res2Net: A New Multi-Scale Backbone Architecture. IEEE Transactions on Pattern Analysis and Machine In-telligence, 43, 652-662.
[17] Heimann, T., Ginneken, B.V., Styner, M.A., Arzhaeva, Y. and Wolf, I. (2009) Comparison and Evaluation of Methods for Liver Segmentation from CT Datasets. IEEE Transactions on Medical Imaging, 28, 1251-1265. [Google Scholar] [CrossRef] 
[18] Liu, S.T., Huang, D. and Wang, Y.H. (2018) Receptive Field Block Net for Accurate and Fast Object Detection. European Conference on Computer Vision, Munich, 8-14 September 2018, 409-419. [Google Scholar] [CrossRef] 
[19] Bilic, P., Christ, P.F., Vorontsov, E., Chlebus, G. and Wu, J. (2019) The Liver Tumor Segmentation Benchmark (LiTS).
https://arxiv.org/abs/1901.04056v1
[20] Liu, T., Liu, J., Ma, Y., He, J., Han, J., Ding, X., et al. (2020) Spatial Feature Fusion Convolutional Network for Liver and Liver Tumor Segmentation from CT Images. Medical Physics, 48, 264-272.