融合Transformer和级联聚合模块的细胞分割算法

doi:10.3778/j.issn.1002-8331.2404-0311

摘要/Abstract

摘要： 针对细胞形态不规则、大小差异大、目标区域密度高和目标区域受到遮挡干扰等问题，提出一种融合Transformer和级联聚合模块肠镜活检组织细胞分割算法。运用金字塔视觉变压器PVTv2作为主干特征提取网络，逐层地提取细胞图像的空间细节和语义信息，构建特征信息之间的长期交互。设计级联聚合模块，不断将高维信息向低维映射来完成全局空间信息与语义信息的融合，同时依靠输入依赖的深度卷积来聚合局部信息，完成特征信息的全面融合。构造混合压缩注意力机制，减少通道维度特征信息冗余并聚焦有效信息，增强空间感知能力，有效减少遮挡和噪声的干扰。在EBHI-Seg数据集上进行实验，并在其子数据集Low-grade IN和Serratedadenoma上取得最佳结果，Dice相似系数和Jaccard指数分别达到93.00%、93.75%和87.21%、88.44%。实验结果表明，该算法能够有效解决细胞分割中被遮挡问题和目标区域细胞过于密集等问题，分割性能较已有算法有所提升。

关键词: 肠镜活检组织细胞, Transformer, 级联聚合模块, 混合压缩注意力机制

Abstract: A cell segmentation algorithm for colonoscopy biopsy tissue that integrates Transformer and cascading aggregation module is proposed to address issues such as irregular cell morphology, large size differences, high density of the target area, and the interference of the target area by occlusion. Firstly, the algorithm uses the pyramid vision transformer PVTv2 as the backbone feature extraction network to extract spatial details and semantic information of cell images layer by layer, and constructs long-term interaction between feature information. Secondly, with cascading aggregation modules, high-dimensional information is continuously mapped to low dimensional information to achieve the fusion of global spatial information and semantic information. At the same time, input dependent deep convolution modules are used to aggregate local information and complete the comprehensive fusion of feature information. Finally, by utilizing a hybrid compressed attention mechanism, the redundancy in channel dimension feature information is reduced while focusing on effective information, enhancing spatial perception ability and effectively reducing the interference of occlusion and noise. Experiments are conducted on the EBHI-Seg dataset, and the best results are obtained on its two subsets Low-grade IN and Serratedadenoma, with Dice similarity coefficient and Jaccard index reaching 93.00%, 93.75%, 87.21% and 88.44% respectively. The experimental results show that the algorithm can effectively solve the problems of occlusion in cell segmentation and excessive cell density in the target area, and the segmentation performance has been improved compared to existing algorithms.

Key words: colonoscopy biopsy tissue cells, Transformer, cascading aggregation module, hybrid compressed attention mechanism

杨国亮, 王乾琛, 耿珍, 熊文楷. 融合Transformer和级联聚合模块的细胞分割算法[J]. 计算机工程与应用, 2025, 61(17): 292-303.

YANG Guoliang, WANG Qianchen, GENG Zhen, XIONG Wenkai. Cell Segmentation Algorithm Integrating Transformer and Cascading Aggregation Module[J]. Computer Engineering and Applications, 2025, 61(17): 292-303.

参考文献

[1] SUNG H, FERLAY J, SIEGEL R L, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA: A Cancer Journal for Clinicians?, 2021, 71(3): 209-249.
[2] 孙福艳, 王琼, 吕宗旺, 等.深度学习在结肠息肉分割中的应用综述[J].计算机工程与应用, 2023, 59(23): 15-27.
SUN F Y, WANG Q, LYU Z W, et al. Review of application of deep learning in colon polyp segmentation[J]. Computer Engineering and Applications, 2023, 59(23): 15-27.
[3] LEE J W J, ZHU F, SRIVASTAVA S, et al. Severity of gastric intestinal metaplasia predicts the risk of gastric cancer: a prospective multicentre cohort study (GCEP)[J]. Gut, 2022, 71(5): 854-863.
[4] 崔浩阳, 丁偕, 张敬谊. 基于细胞图卷积的组织病理图像分类研究[J]. 计算机工程与应用, 2020, 56(24): 223-228.
CUI H Y, DING X, ZHANG J Y. Research on classification of histopathological image based on cell graph convolutional network[J]. Computer Engineering and Applications, 2020, 56(24): 223-228.
[5] LI Y, ZHU R, MI L, et al. Segmentation of white blood cell from acute lymphoblastic leukemia images using dual-threshold method[J].Computational and Mathematical Methods in Medicine, 2016, 2016: 9514707.
[6] SZENASI S. Distributed region growing algorithm for medical image segmentation[J]. International Journal of Circuits, Systems and Signal Processing, 2014, 8(1): 173-181.
[7] MOLNAR C, JERMYN I H, KATO Z, et al. Accurate morphology preserving segmentation of overlapping cells based on active contours[J].Scientific Reports, 2016, 6: 32412.
[8] LIN C H, CHEN C C. Image segmentation based on edge dete-ction and region growing for thinprep-cervical smear[J]. International Journal of Pattern Recognition and Artificial Intelligence, 2010, 24(7): 1061-1089.
[9] ZHANG C, YARKONY J, HAMPRECHT F A. Cell detection and segmentation using correlation clustering[C]//Proceedings of the Medical Image Computing and Computer-Assisted Intervention. Cham: Springer International Publishing, 2014: 9-16.
[10] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]//Proceedings of the Medical Image Computing and Computer-Assisted Intervention. Cham: Springer International Publishing, 2015: 234-241.
[11] JOSEPH N, KOLLURU C, BENETZ B A M, et al. Quantitative and qualitative evaluation of deep learning automatic segmentations of corneal endothelial cell images of reduced image quality obtained following cornea transplant[J]. J Med Imaging (Bellingham), 2020, 7(1): 014503.
[12] ROY M R, AMEER P M. Segmentation of leukocyte by semantic segmentation model: a deep learning approach[J]. Biomedical Signal Processing and Control, 2021, 65: 102385.
[13] CHANCHAL A K, KUMAR A, LAL S, et al. Efficient and robust deep learning architecture for segmentation of kidney and breast histopathology images[J]. Computers & Electrical Engineering, 2021, 92: 107177.
[14] FIAZ M, HEIDARI M, ANWER R, et al. SA2-Net: scale-aware attention network for microscopic image segmentation[J]. arXiv: 2309.16661, 2023.
[15] TANG F L, XU Z X, HUANG Q M, et al. DuAT: dual-aggregation transformer network forMedical image segmentation[C]//Proceedings of the Pattern Recognition and Computer Vision.Singapore: Springer Nature Singapore, 2024: 343-356.
[16] WANG H N, CAO P, LIU X L, et al. Narrowing the semantic gaps in U-Net with learnable skip connections: the case of medical image segmentation[J]. arXiv:2312.15182, 2023.
[17] WANG W H, XIE E Z, LI X, et al. PVT v2: improved baselines with pyramid vision transformer[J]. Computational Visual Media, 2022, 8(3): 415-424.
[18] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16×16 words: Transformers for image recognition at scale[J]. arXiv:2010.11929, 2020.
[19] LOU M, ZHANG S, ZHOU H Y, et al. TransXNet: learning both global and local dynamics with a dual dynamic token mixer for visual recognition[J]. arXiv:2310.19380, 2023.
[20] PARK J, WOO S, LEE J Y, et al. BAM: bottleneck attention module[J].arXiv:1807.06514, 2018.
[21] WEI J, WANG S H, HUANG Q M. F3Net: fusion, feedback and focus for salient object detection[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 12321-12328.
[22] SHI L, LI X, HU W, et al. EBHI-Seg: a novel enteroscope biopsy histopathological hematoxylin and eosin image dataset for image segmentation tasks[J]. Front Med (Lausanne), 2023, 10: 1114673.
[23] NAYLOR P, LAé M, REYAL F, et al. Segmentation of nuclei in histopathology images by deep regression of the distance map[J]. IEEE Transactions on Medical Imaging, 2019, 38(2): 448-459.
[24] IQBAL A, SHARIF M, KHAN M A, et al. FF-UNet: a U-shaped deep convolutional neural network for multimodal biomedical image segmentation[J]. Cognitive Computation, 2022, 14(4): 1287-1302.
[25] FAN D P, JI G P, ZHOU T, et al. PraNet: parallel reverse atte-ntion network for polyp segmentation[C]//Proceedings of the Medical Image Computing and Computer Assisted Intervention. Cham: Springer International Publishing, 2020: 263-273.
[26] WANG J F, HUANG Q M, TANG F L, et al. Stepwise feature fusion: local guides global[C]//Proceedings of the Medical Image Computing and Computer Assisted Intervention. Cham: Springer Nature Switzerland, 2022: 110-120.
[27] WU C, LONG C, LI S J, et al. MSRAformer: multiscale spatial reverse attention network for polyp segmentation[J]. Computers in Biology and Medicine, 2022, 151: 106274.
[28] METLEK S. CellSegUNet: an improved deep segmentation model for the cell segmentation based on UNet++?and residual UNet models[J]. Neural Computing and Applications, 2024, 36(11): 5799-5825.
[29] MONDAL S K, Talukder M S H , ALJAIDI M, et al. BloodCell-Net: a lightweight convolutional neural network for the classification of all microscopic blood cell images of the human body[J]. arXiv:2405.14875, 2024.