外周血白细胞在人体免疫系统中起着不可忽视的作用，对医生进行病情诊断有重要价值。目前，医院主要采用人工镜检与血细胞分析仪检测白细胞，存在工作量大、耗时长、统计量小、易受主观因素影响等问题。因此，实现精准白细胞分类识别有着迫切的现实需求。

       针对白细胞外观复杂、目标尺寸小而导致识别精确度低、效果不佳等问题，本文基于YOLOv5网络模型，提出了一种改进YOLOv5的白细胞识别算法。首先在主干网络的卷积层中添加坐标注意力机制，将位置信息嵌入到通道注意力中以提升网络的特征提取能力；其次采用多尺度特征检测层，充分利用浅层特征信息以减少卷积过程中小目标区域特征信息的丢失，提高了小尺寸白细胞的识别精度。为了解决白细胞分类识别任务中类别不平衡的问题，在改进YOLOv5白细胞识别算法的基础上，提出一种基于YOLOv5-CHCE的不平衡白细胞识别算法。该算法使用类平衡焦点损失函数作为网络的分类损失函数，重新平衡了各类白细胞的损失权重，以提升不平衡白细胞数据集中少数类的识别率；同时，采用有效交并比损失函数作为网络的边框回归损失函数，提升检验框识别的精确度。

       本文使用公共白细胞数据集和自建的SWBC（Self-built White Blood Cells）数据集分别对改进算法进行实验评估。实验结果表明，改进YOLOv5-CHCE模型在类别不平衡的公共白细胞数据集中精确度、召回率和平均精确度均值（mean Average Precision，mAP）分别达到了98.7%、98.5%和98.5%；在自建SWBC数据集中精确度、召回率和mAP分别达到了93%、94.3%和97.1%。改进的网络模型在不同的数据集上展现出优越性，实现了更高精度的多类白细胞分类识别。

     Peripheral blood white blood cells play a role that cannot be ignored in the human immune system and are of great value to doctors in diagnosing the disease. At present, hospitals mainly use manual microscopy and cell analyzers to detect white blood cells, but there are problems such as heavy workload, long time, small statistics, and being easily affected by subjective factors. Therefore, there is an urgent practical need to achieve accurate white blood cells classification and identification.

     In view of the complex appearance and small size of white blood cells, which lead to low recognition accuracy and poor effect, this paper proposes an improved YOLOv5 white blood cells recognition algorithm based on the YOLOv5 network model. First, a coordinate attention mechanism is added to the convolutional layer of the backbone network, which embeds the position information into the channel attention to improve the feature extraction ability of the network. Then, the multi-scale feature detection layers are adopted to make full use of the shallow feature information to reduce the loss of feature information of small target areas during the convolution process, and to improve the identification accuracy of white blood cells with smaller size. In order to solve the problem of class imbalance in the task of white blood cells classification and identification, an unbalanced white blood cells identification algorithm based on YOLOv5-CHCE is proposed on the basis of improving the YOLOv5 white blood cells identification algorithm. The algorithm uses the class-balanced focal loss function as the classification loss function of the network, which rebalances the loss weights of various types of white blood cells to improve the recognition rate of minority classes in the imbalanced white blood cells dataset. At the same time, the effective intersection ratio loss function is used to improve the frame regression loss function of the network to improve the accuracy of the detection frame recognition.

     This paper uses the public white blood cells dataset and the self-built SWBC (Self-built White Blood Cells) dataset to conduct experimental evaluations on the improved algorithm. The experimental results show that the improved YOLOv5-CHCE model proposed in this paper achieves 98.7%, 98.5% and 98.5% of the precision, recall and mean Average Precision (mAP) in the class-imbalanced public white blood cells dataset, respectively. In the self-built SWBC dataset, the precision, recall and mAP reached 93%, 94.3% and 97.1%, respectively. The improved network model shows superiority in different datasets and achieves higher-precision classification and identification of multiple types of white blood cells.

[1]王卫星,苏培垠.基于颜色、梯度矢量流活动轮廓及支持向量机实现白细胞的提取和分类[J].光学精密工程,2012,20(12):2781-2790.

[2]吴上上.两级目标检测器白细胞自动分类与识别技术[D].济南:山东师范大学,2021.

[3]李国权,姚凯,庞宇.基于CNN的血液细胞图像自动识别算法[J].计算机科学,2022,49(04):247-253.

[4]Chen W, Yang X. Dimension Detection of Varistor Based on Random Hough Transform[C]// The 2nd International Conference on Sensor Networks and Signal Processing (SNSP 2019). Taiwan, China, 2019:117-126.

[5]陈振华,杨茂红.浅谈TEK-Ⅱ血细胞分析仪原理及故障维修[J].中国设备工程,2021(15)：50-51.

[6]段浩,陈锋,顾彪,李抄,吴太虎.血细胞分析技术及其进展研究[J].医疗卫生装备,2014,35(05):108-112.

[7]Gayathri. S , Jyothi R L. An Automated Leucocyte Classification For Leukemia Detection[J]. International Research Journal of Engineering and Technology (IRJET).2018, 5(5):4254-4264.

[8]Patil S, Patane P P R S, Patil M. Acute Lymphoblastic LeukemiaDetectionin Human Blood using Microscopic Image[J]. Mukt Shabd Journal,2020,13:1539-1544.

[9]Sahlol A T, Kollmannsberger P, Ewees A A. Efficient classification of white blood cell leukemia with improved swarm optimization of deep features[J]. Scientific Reports, 2020, 10(1): 1-11.

[10]刘霖,于腾,李佼,傅小明等.一种基于机器视觉的全自动血涂片形态分析装置[P].中国,202010261704.X,2020年4月4日.

[11]孙凯,姚旭峰,黄钢.基于机器学习的白细胞六分类研究[J].软件,2020,41(10):98-101+134.

[12]张剑飞,郭笑颜,王波,崔文升.基于迁移学习和支持向量机的白细胞分类[J].科学技术与工程,2021,21(19):8113-8119.

[13]Jung C, Abuhamad M, Alikhanov J, et al. W-net: a CNN-based architecture for white blood cells image classification[J]. Electrical Engineering and Systems Science,2019,10:11-16.

[14]王亚品,曹益平,付光凯,王璐,万莹莹,李城梦.基于深度卷积神经网络的人体外周血白细胞显微图像分类[J].光电子·激光,2019,30(05):546-555.

[15]Basnet J, Alsadoon A, Prasad P W C, et al. A novel solution of using deep learning for white blood cells classification: enhanced loss function with regularization and weighted loss (ELFRWL)[J]. Neural Processing Letters, 2020, 52(2): 1517-1553.

[16]Li J , Wu J M. Leukocyte Detection in Blood Smear Image Based on Improved YOLOv3[J].Proceedings of 2020 the 10th International Workshop on Computer Science and Engineering , 2020,14: 144-149.

[17]Redmon J, Farhadi A. YOLOv3: An Incremental Improvement[C]//IEEE Conference on Computer Vision and Pattern Recognition. USA,IEEE, 2018:2311-2314.

[18]马力,帅仁俊,刘文佳,李文煜.基于改进的残差神经网络的白细胞分类[J].计算机工程与设计,2020,41(10):2982-2987.

[19]Baghel N, Verma U, Nagwanshi K K. WBCs-Net: type identification of white blood cells using convolutional neural network[J]. Multimedia Tools and Applications, 2021: 1-17.

[20]耿磊,杨盟盟,肖志涛,张芳,刘彦北,吴骏,王雯.结合通道空间加权特征金字塔网络的白细胞检测与分割[J].计算机辅助设计与图形学学报,2021,33(09):1418-1427.

[21]Yao J, Huang X, Wei M, et al. High-Efficiency Classification of White Blood Cells Based on Object Detection[J]. Journal of Healthcare Engineering, 2021,(2021):1615192-1615192.

[22]Bochkovskiy A, Wang C Y, Liao H Y M. Yolov4: Optimal speed and accuracy of object detection[J]. Computer Vision and Pattern Recognition,2020,4:1-17.

[23]Ren S, He K, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 28(6): 1137-1149.

[24]Wang D, Hwang M, Jiang W C, et al. A deep learning method for counting white blood cells in bone marrow images[J]. BMC BIOINFORMATICS, 2021, 22(5): 1-14.

[25]Lin T Y, Dollár P, Girshick R, et al. Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 2117-2125.

[26]杨勇超.基于卷积神经网络的白细胞分类识别方法研究[D].西安:西安科技大学,2021.

[27]Cheuque C, Querales M, León R, et al. An Efficient Multi-Level Convolutional Neural Network Approach for White Blood Cells Classification[J]. Diagnostics, 2022, 12(2): 248.

[28]Lin T Y, Goyal P, Girshick R, et al. Focal Loss for dense object detection[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2980-2988.

[29]Pouyanfar S, Tao Y, Mohan A, et al. Dynamic sampling in convolutional neural networks for imbalanced data classification[C]//2018 IEEE conference on multimedia information processing and retrieval (MIPR). IEEE, 2018: 112-117.

[30]Cui Y, Jia M, Lin T Y, et al. Class-balanced loss based on effective number of samples[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 9268-9277.

[31]Cao K,Wei C, Gaidon A, et al. Learning imbalanced datasets with label-distribution-aware margin loss[J]. Advances in neural information processing systems, 2019, 32.

[32]Puri A, Kumar Gupta M. Improved hybrid bag-boost ensemble with K-means-SMOTE–ENN technique for handling noisy class imbalanced data[J]. The Computer Journal, 2022, 65(1): 124-138.

[33]Wang S, Liu W, Wu J, et al. Training deep neural networks on imbalanced data sets[C]//2016 international joint conference on neural networks (IJCNN). IEEE, 2016: 4368-4374.

[34]Huang C, Li Y, Change Loy C, et al. Learning deep representation for imbalanced classification[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 5375-5384.

[35]Lin W C, Tsai C F, Hu Y H, et al. Clustering-based undersampling in class-imbalanced data[J]. Information Sciences, 2017, 409: 17-26.

[36]Wang H, Cui Z, Chen Y, et al. Predicting hospital readmission via cost-sensitive deep learning[J]. IEEE/ACM transactions on computational biology and bioinformatics, 2018, 15(6): 1968-1978.

[37]Dong Q, Gong S, Zhu X. Imbalanced deep learning by minority class incremental rectification[J]. IEEE transactions on pattern analysis and machine intelligence, 2018, 41(6): 1367-1381.

[38]张家伟,郭林明,杨晓梅.针对不平衡数据的过采样和随机森林改进算法[J].计算机工程与应用,2020,56(11):39-45.

[39]王德志,梁俊艳.不平衡数据集文本多分类深度学习算法[J].计算机工程与设计,2021,42(09):2501-2508.

[40]Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2014: 580-587.

[41]He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2015, 37(9): 1904-1916.

[42]Girshick R. Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2015: 1440-1448.

[43]Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788.

[44]Liu W, Anguelov D, Erhan D, et al. Ssd: Single shot multibox detector[C]//European conference on computer vision. Springer, Cham, 2016: 21-37.

[45]Redmon J, Farhadi A. YOLO9000: better, faster, stronger[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 7263-7271.

[46]Wang W, Xie E, Song X, et al. Efficient and accurate arbitrary-shaped text detection with pixel aggregation network[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019: 8440-8449.

[47]Wang C Y, Liao H Y M, Wu Y H, et al. CSPNet: A new backbone that can enhance learning capability of CNN[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops. 2020: 390-391.

[48]He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2015, 37(9): 1904-1916.

[49]Zheng Z, Wang P, Liu W, et al. Distance-IoU loss: Faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2020, 34(07): 12993-13000.

[50]Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141.

[51]Hou Q, Zhou D, Feng J. Coordinate attention for efficient mobile network design[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2021: 13713-13722.

[52]Rezatofighi S H , Soltanian-Zadeh H . Automatic recognition of five types of white blood cells in peripheral blood[J]. Computerized Medical Imaging & Graphics, 2011, 35(4):333-343.

[53]Shenggan,“BCCD dataset,”2017, https://github.com/Shenggan/BCCD Dataset, Accessed 01 August 2018.

[54]回天,哈力旦·阿布都热依木,杜晗.结合Faster R-CNN的多类型火焰检测[J].中国图象图形学报,2019,24(01):73-83.

[55]Zhao J,Zhang M,Zhou Z.Automatic detection and classification of leukocytes using convolutional neural networks[J].Medical & biological engineering & computing,2017,55(8):1287-1301.

[56]Zhang Y F, Ren W, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. Computer Science,2021.