广西师范大学学报（自然科学版） ›› 2023, Vol. 41 ›› Issue (3): 67-79.doi: 10.16088/j.issn.1001-6600.2022101301

• 研究论文 • 上一篇    下一篇

基于双编码路径融合和双向ConvLSTM的神经元图像分割

钱有为1, 何富运1,2*, 韦燕1, 冯慧玲1, 胡聪2   

  1. 1.广西师范大学 电子与信息工程学院, 广西 桂林 541004;
    2.广西自动检测技术与仪器重点实验室(桂林电子科技大学), 广西 桂林 541004
  • 收稿日期:2022-10-13 修回日期:2022-12-26 出版日期:2023-05-25 发布日期:2023-06-01
  • 通讯作者: 何富运(1982—), 男, 广西陆川人, 广西师范大学副教授, 博士。E-mail: he_fuyun@gxnu.edu.cn
  • 基金资助:
    国家自然科学基金(62062014); 广西自然科学基金(2018GXNSFAA050024); 广西师范大学重点科学研究计划(2018ZD007)

Neuron Image Segmentation Based on Dual Coding Path Fusion and Bidirectional ConvLSTM

QIAN Youwei1, HE Fuyun1,2*, WEI Yan1, FENG Huiling1, HU Cong2   

  1. 1. School of Electronic and Information Engineering, Guangxi Normal University, Guilin Guangxi 541004, China;
    2. Guangxi Key Laboratory of Automatic Detecting Technology and Instruments (Guilin University of Electronic Technology), Guilin Guangxi 541004, China
  • Received:2022-10-13 Revised:2022-12-26 Online:2023-05-25 Published:2023-06-01

PDF (PC)

62

摘要： 目标分割是神经元图像分析中必不可少的步骤之一,分割的准确性会直接影响到神经元图像分析和重建的质量。在面对结构边界模糊、存在弱噪声或弱纤维信号的神经元图像时,已有的分割方法依然存在误差较大、识别信号不准等问题。为了解决这些问题,基于神经元的特征,本文提出一种基于双编码路径融合和双向 ConvLSTM的深度学习网络(DFC-Net)用于神经元图像分割。首先,网络在编码器阶段采用双编码路径提取特征,其中第一路编码路径采用基于空洞卷积的密集连接网络作为固定特征提取器,第二路编码器采用深度残差网络作为特征提取网络;接着,使用密集连接ASPP网络作为桥梁连接编码器和解码器;最后,在跳跃连接中使用双向ConvLSTM结合编码器和解码器,在解码器阶段引入融合网络以融合2个编码器提取的特征,从而增强空间信息的传播。多组对比实验结果显示,本文提出的网络有效地提高了电子显微镜神经元图像的分割精度,在ISBI-2012和SNEMI3D数据集上的Sen、Dice分别达到0.952 7、0.958 9和0.941 6、0.912 7,平均准确率相比于其他U-Net变体网络提高2.93%。

关键词: 图像分割, 神经元, 双编码路径, D-ASPP, 双向ConvLSTM

Abstract: Target segmentation is one of the essential steps in neuronal image analysis. The accuracy of segmentation directly affects the quality of neuronal image analysis and reconstruction. In the face of fuzzy structure boundary, weak noise or weak fiber signal, the existing segmentation methods still have some problems, such as large error and inaccurate recognition signal. To solve these problems, based on the characteristics of neurons, a deep learning network based on dual coding path fusion and bidirectional ConvLSTM is proposed for neuronal image segmentation. Firstly, the network adopts dual coding paths to extract features in the encoder stage, in which the first encoding path adopts dense connection network based on dilated convolution as the fixed feature extractor, and the second encoder adopts deep residual network as the feature extraction network. Then, the densely connected ASPP network is used as a bridge to connect the encoder and decoder. Finally, bidirectional ConvLSTM is used to combine encoder and decoder in skip connection, and a fusion network is introduced in the decoder stage to fuse the features extracted by the two encoders, so as to enhance the spatial information propagation. The results of multiple comparative experiments show that the proposed network can effectively improve the segmentation accuracy of electron microscope neuron images. Sen and Dice on ISBI-2012 and SNEMI3D datasets reaches 0.952 7, 0.958 9 and 0.941 6, 0.912 7, respectively, and the average accuracy is higher than that of U-net and other models.

Key words: image segmentation, neuron, dual coding path, D-ASPP, bidirectional ConvLSTM

中图分类号:  TP391.41

[1] BOUÇA-MACHADO R, TITOVA N, CHAUDHURI K R, et al. Palliative care for patients and families with Parkinson's disease[J]. International Review of Neurobiology, 2017, 132: 475-509. DOI: 10.1016/bs.irn.2017.02.017.
[2] HU J M, LIU S Y. Modulating intracellular oxidative stress via engineered nanotherapeutics[J]. Journal of Controlled Release, 2020, 319: 333-343. DOI: 10.1016/j.jconrel.2019.12.040.
[3] ZHANG Q R, ZHANG M, CHEN T H, et al. Recent advances in convolutional neural network acceleration[J]. Neurocomputing, 2019, 323: 37-51. DOI: 10.1016/j.neucom.2018.09.038.
[4] LU Y, LU G M, LI J X, et al. High-parameter-efficiency convolutional neural networks[J]. Neural Computing and Applications, 2020, 32(14): 10633-10644. DOI: 10.1007/s00521-019-04596-w.
[5] 张鑫, 姚庆安, 赵健, 等. 全卷积神经网络图像语义分割方法综述[J]. 计算机工程与应用, 2022, 58(8): 45-57. DOI: 10.3778/j.issn.1002-8331.2109-0091.
[6] SHELHAMER E, LONG J, DARRELL T. Fully convolutional networks for semantic segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(4): 640-651. DOI: 10.1109/tpami.2016.2572683.
[7] RONNEBERGER O, FISCHER P, BROX T. U-Net: Convolutional networks for biomedical image segmentation[C]//Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015: LNCS Volume 9351. Cham: Springer, 2015: 234-241. DOI: 10.1007/978-3-319-24574-4_28.
[8] ZHOU Z W, RAHMAN SIDDIQUEE M M, TAJBAKHSH N, et al. UNet++: a nested U-Net architecture for medical image segmentation[C]//Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: LNCS Volume 11045. Cham: Springer, 2018: 3-11. DOI: 10.1007/978-3-030-00889-5_1.
[9] 吴相远, 申诺, 蒙玉洪, 等.基于U-Net++的人体脊椎MRI图像识别[J]. 桂林电子科技大学学报, 2022, 42(1): 36-42. DOI: 10.16725/j.cnki.cn45-1351/tn.2022.01.004.
[10] 徐思则, 刘威. 基于UNet网络的乳腺癌肿瘤细胞图像分割[J]. 电子设计工程, 2022, 30(12): 63-66, 73. DOI: 10.14022/j.issn1674-6236.2022.12.013.
[11] 张秦瑞, 林国军, 朱晏梅. 改进SegNet网络对遥感图像的语义分割[J]. 宜宾学院学报, 2022, 22(6): 37-41. DOI: 10.19504/j.cnki.issn1671-5365.2022.06.08.
[12] 雷竞雄. 基于改进的SegNet分割网络的遥感图像分割[J]. 理论数学, 2022, 12(11): 1875-1881. DOI: 10.12677/PM.2022.1211201.
[13] GAO S H, CHENG M M, ZHAO K, et al. Res2Net: a new multi-scale backbone architecture[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(2): 652-662. DOI: 10.1109/TPAMI.2019.2938758.
[14] 高慧, 阎晓东, 张衡, 等. 基于Res2Net的多尺度遥感影像海陆分割方法[J]. 光学学报, 2022, 42(18): 147-154. DOI: 10.3788/AOS202242.1828004.
[15] 刘尚旺, 张杨杨, 蔡同波, 等. 基于改进PSPnet的无人机农田场景语义分割[J]. 灌溉排水学报, 2022, 41(4): 101-108. DOI: 10.13522/j.cnki.ggps.2021406.
[16] ZHAO H S, SHI J P, QI X J, et al. Pyramid scene parsing network[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Los Alamitos, CA: IEEE Computer Society, 2017: 6230-6239. DOI: 10.1109/CVPR.2017. 660.
[17] CHEN L C, PAPANDREOU G, KOKKINOS I, et al. DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4): 834-848. DOI: 10.1109/TPAMI.2017.2699184.
[18] LI J, OU X, SHEN N Y, et al. Study on strategy of CT image sequence segmentation for liver and tumor based on U-Net and Bi-ConvLSTM[J]. Expert Systems with Applications, 2021, 180: 115008. DOI: 10.1016/j.eswa.2021.115008.
[19] 司明明, 陈玮, 胡春燕, 等. 融合Resnet50和U-Net的眼底彩色血管图像分割[J]. 电子科技, 2021, 34(8): 19-24. DOI: 10.16180/j.cnki.issn1007-7820.2021.08.004.
[20] 秦志远, 朱峻泷, 张琛, 等. 基于ResNet34-UNet的静脉超声图像分割方法研究[J]. 临床超声医学杂志, 2022, 24(1): 74-78. DOI: 10.16245/j.cnki.issn1008-6978.2022.01.001.
[21] 王卓英, 童基均, 蒋路茸, 等. 基于U-Dense-net网络的DSA图像冠状动脉血管分割[J]. 浙江理工大学学报(自然科学版), 2021, 45(3): 390-399. DOI: 10.3969/j.issn.1673-3851(n).2021.03.013.
[22] 魏榕剑, 邵剑飞. 基于改进Dense Net网络的肺炎X光图像识别算法[J]. 电视技术, 2021, 45(6): 140-143. DOI: 10.16280/j.videoe.2021.06.040.
[23] OFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on Machine Learning. Lille: PMLR, 2015: 448-456.
[24] CHENG J L, TIAN S W, YU L, et al. ResGANet: residual group attention network for medical image classification and segmentation[J]. Medical Image Analysis, 2022, 76: 102313. DOI: 10.1016/j.media.2021.102313.
[25] HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Los Alamitos, CA: IEEE Computer Society, 2017: 2261-2269. DOI: 10.1109/CVPR.2017.243.
[26] DUTTA K. Densely connected recurrent residual (dense R2UNet) convolutional neural network for segmentation of lung CT images[EB/OL]. (2021-02-01)[2022-10-13]. https://arxiv.org/abs/2102.00663. DOI: 10.48550/arXiv.2102.00663.
[27] ZHOU E J, XU X, XU B M, et al. An enhancement model based on dense atrous and inception convolution for image semantic segmentation[J]. Applied Intelligence,2023,53(5):5519-5531. DOI: 10.1007/s10489-022-03448-w.
[28] YANG B, CHEN W X, LUO H Q, et al. Neuron image segmentation via learning deep features and enhancing weak neuronal structures[J]. IEEE Journal of Biomedical and Health Informatics, 2021, 25(5): 1634-1645. DOI: 10.1109/jbhi.2020.3017540.
[29] YANG B, LIU M, WANG Y N, et al. Structure-guided segmentation for 3D neuron reconstruction[J]. IEEE Transactions on Medical Imaging, 2021, 41(4): 903-914. DOI: 10.1109/tmi.2021.3125777.
[30] JIANG Y, CHEN W X, LIU M, et al. 3D neuron microscopy image segmentation via the ray-shooting model and a DC-BLSTM network[J]. IEEE Transactions on Medical Imaging, 2021, 40(1): 26-37. DOI: 10.1109/tmi.2020.3021493.
[31] WANG H, ZHANG D H, SONG Y, et al. Segmenting neuronal structure in 3D optical microscope images via knowledge distillation with teacher-student network[C]//2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019). Piscataway NJ: IEEE, 2019: 228-231. DOI: 10.1109/ISBI.2019.8759326.
[32] HU J, SHEN L, ALBANIE S, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(8): 2011-2023. DOI: 10.1109/tpami.2019.2913372.
[33] CHEN L C, PAPANDREOU G, KOKKINOS I, et al. Semantic image segmentation with deep convolutional nets and fully connected CRFs[EB/OL]. (2014-12-22)[2022-10-13]. https://arxiv.org/abs/1412.7062v1. DOI: 10.48550/arXiv.1412.7062.
[34] WANG Z M, WANG J S, YANG K, et al. Semantic segmentation of high-resolution remote sensing images based on a class feature attention mechanism fused with Deeplabv3+[J]. Computers & Geosciences, 2022, 158: 104969. DOI: 10.1016/j.cageo.2021.104969.
[35] CARDONA A, SAALFELD S, PREIBISCH S, et al. An integrated micro-and macro architectural analysis of the drosophila brain by computer-assisted serial section electron microscopy[J]. PLoS Biology, 2010, 8(10): e1000502. DOI: 10.1371/journal.pbio.1000502.
[36] LEE K, ZUNG J, LI P, et al. Superhuman accuracy on the SNEMI3D connectomics challenge[EB/OL]. (2017-05-31)[2022-10-13]. https://arxiv.org/abs/1706.00120v1. DOI: 10.48550/arXiv.1706.00120.
[37] 丁才富, 杨晨, 纪秋浪, 等. MCA-Net: 多尺度综合注意力CNN在医学图像分割中的应用[J]. 微电子学与计算机, 2022, 39(3): 71-77. DOI: 10.19304/j.issn1000-7180.2021.0950.
[38] 李玉慧, 梁创学, 李军. 基于Group-Depth U-Net的电子显微图像中神经元结构分割[J]. 中国医学物理学杂志, 2020, 37(6): 720-725. DOI: 10.3969/j.issn.1005-202X.2020.06.012.
[39] ROY S, CHAKRABORTY A. Semantic segmentation of highly class imbalanced fully labeled 3D volumetric biomedical images and unsupervised domain adaptation of the pre-trained segmentation network to segment another fully unlabelled biomedical 3D image stack[EB/OL]. (2020-03-13)[2022-10-13]. https://arxiv.org/abs/2004.02748. DOI: 10.48550/arXiv.2004.02748.
[40] URAKUBO H, BULLMANN T, KUBOTA Y, et al. UNI-EM: an environment for deep neural network-based automated segmentation of neuronal electron microscopic images[J]. Scientific Reports, 2019, 9: 19413. DOI: 10.1038/s41598-019-55431-0.
[1] 马军. 功能神经元建模及动力学若干问题[J]. 广西师范大学学报（自然科学版）, 2022, 40(5): 307-323.
[2] 万黎明, 张小乾, 刘知贵, 宋林, 周莹, 李理. 基于高效通道注意力的UNet肺结节CT图像分割[J]. 广西师范大学学报（自然科学版）, 2022, 40(3): 66-75.
[3] 谭安杰, 韦笃取, 覃英华. 电磁场耦合忆阻神经网络的放电模式及同步行为研究[J]. 广西师范大学学报（自然科学版）, 2020, 38(1): 107-113.
[4] 王勋, 李廷会, 潘骁, 田宇. 基于改进模糊C均值聚类与Otsu的图像分割方法[J]. 广西师范大学学报（自然科学版）, 2019, 37(4): 68-73.
[5] 夏海英. 基于改进的SLIC区域合并的宫颈细胞图像分割[J]. 广西师范大学学报（自然科学版）, 2016, 34(4): 93-100.
[6] 张新明, 张玉珊, 李振云. 一种改进的矩不变图像分割方法[J]. 广西师范大学学报（自然科学版）, 2011, 29(2): 185-190.
[7] 冯嘉礼, 杨润泽. 属性论方法在图像分割中的应用研究[J]. 广西师范大学学报（自然科学版）, 2011, 29(2): 191-194.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
[1] 陈泽柠, 罗雪梅, 廖娴, 陈郭燕, 武正军. 氯化钙对福寿螺卵块的杀灭作用[J]. 广西师范大学学报（自然科学版）, 2016, 34(1): 156 -161 .
[2] 冉浩, 周善义. 中国蚁科昆虫名录——蚁型亚科群(膜翅目:蚁科)(Ⅲ)[J]. 广西师范大学学报（自然科学版）, 2013, 31(1): 104 -111 .
[3] 杜雪松,林勇,梁国琨,黄姻,宾石玉,陈忠,覃俊奇,赵怡. 两种罗非鱼的耐寒性能比较[J]. 广西师范大学学报（自然科学版）, 2019, 37(3): 174 -179 .
[4] 宾石玉, 廖芳, 杜雪松, 许艺兰, 王鑫, 武霞, 林勇. 罗非鱼耐寒性能研究进展[J]. 广西师范大学学报（自然科学版）, 2021, 39(1): 10 -16 .
[5] 陈丹妮, 陈志林, 周善义. 中国蚁科昆虫名录——切叶蚁亚科(补遗)[J]. 广西师范大学学报（自然科学版）, 2021, 39(1): 87 -97 .
[6] 蒋向辉, 谭荣, 杨永平, 肖清淙. 十大功劳甘草汤治疗肝炎的网络药理学研究[J]. 广西师范大学学报（自然科学版）, 2021, 39(5): 198 -209 .
[7] 田镇滔, 张军舰. 基于分位数方法的超高维删失数据的特征筛选[J]. 广西师范大学学报（自然科学版）, 2021, 39(6): 99 -111 .
[8] 曹新光, 岳伟鹏, 邓洁. 北亚热带山地不同海拔土壤有机碳分布特征——以鄂东龟峰山为例[J]. 广西师范大学学报（自然科学版）, 2021, 39(6): 174 -182 .
[9] 杜锦丰, 王海荣, 梁焕, 王栋. 基于表示学习的跨模态检索方法研究进展[J]. 广西师范大学学报（自然科学版）, 2022, 40(3): 1 -12 .
[10] 张萍, 徐巧枝. 基于多感受野与分组混合注意力机制的肺结节分割研究[J]. 广西师范大学学报（自然科学版）, 2022, 40(3): 76 -87 .
版权所有 © 广西师范大学学报（自然科学版）编辑部
地址：广西桂林市三里店育才路15号　邮编：541004
电话：0773-5857325　E-mail: gxsdzkb@mailbox.gxnu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发