作者投稿
专家审稿
编辑办公
邮件订阅
RSS
-
摘要:
目的 利用深度学习技术，建立临床常见的侵袭性真菌图像辅助分类模型。 方法 回顾性收集2020年9月—2021年4月解放军总医院第八医学中心曲霉菌属、酵母菌属和新型隐球菌属真菌感染者的显微镜图像，按7∶1.5∶1.5的比例随机分为训练集、验证集和测试集。使用训练集和验证集图像对改进的MobileNetV2网络结构进行训练和参数调试，构建基于多尺度注意力机制的卷积神经网络(convolutional neural network, CNN)真菌图像11分类模型。以机器鉴定结果为金标准，以查准率、召回率和F1值为指标评价该模型对测试集真菌图像的分类效果。将该模型与5种经典CNN模型进行比较，评价指标包括模型参数量、内存占用量、网络每秒处理的图像数量(frames per second, FPS)、准确率及受试者操作特征曲线下面积(area under the curve，AUC)。 结果 共纳入真菌显微镜图像7666张，分别包括曲霉菌属、酵母菌属和新型隐球菌属图像2781张、4115张、770张。其中训练集5366张、验证集1150张、测试集1150张。改进的MobileNetV2模型对测试集11种真菌图像具有较高的分类性能，查准率为96.36%~100%，召回率为96.53%~100%，F1值为97.01%~100%。该模型的参数量、内存占用量分别为4.22 M、356.89 M，FPS为573，准确率为(99.09±0.18)%，AUC为0.9944±0.0018，综合性能优于5种经典网络模型。 结论 本研究提出的真菌图像分类模型，在保持低运算成本的情况下，可获得较高的真菌图像识别能力，其整体性能优于常见的经典模型。 Abstract:Objective To establish a fungal image-assisted classification model using deep learning technology. Methods The microscope images of people infected with Aspergillus, Saccharomyces and Cryptococcus neoformans were retrospectively collected from the Eighth Medical Center of PLA General Hospital from September 2020 to April 2021. The images were randomly divided into training set, validation set and test set according to the ratio of 7∶1.5∶1.5. The improved MobileNetV2 network structure was trained using the training set, a convolutional neural network (CNN) fungal image 11 classification model based on multi-scale attention mechanism was constructed and the parameters were debugged based on the validation set. Machine identification results were taken as the gold standard, the performance of the model on 11 fungal image classification tasks was evaluated, and the results were shown by precision, recall and F1 value. In addition, the performance of the proposed model with 5 classic CNN models were compared, and the results were measured in terms of model parameters, memory usage, frames per second (FPS), accuracy, and area under the curve (AUC) of receiver operating characteristic curve. Results A total of 7666 fungal microscope images were collected, including 2781, 4115, and 770 images of Aspergillus, Saccharomyces, and Cryptococcus neoformans, respectively. Among them, there were 5366 training images, 1150 validation images, and 1150 test images. The improved MobileNetV2 model had high performance for the classification of 11 fungal images in the test set. The precision rate was distributed between 96.36% and 100%, the recall rate was distributed between 96.53% and 100%, and the F1 value was distributed between 97.01% and 100%. The parameters, memory usage, FPS, accuracy, and AUC of the improved MobileNetV2 model were 4.22 M, 356.89 M, 573, (99.09±0.18)%, and 0.9944±0.0018, respectively, and the comprehensive performance was better than 5 kinds of classical networks. Conclusion The proposed fungal image classification model based on the improved MobileNetV2 can obtain higher fungal image recognition ability while maintaining low computational cost, with an overall performance better than classical CNN model. 作者贡献：张雪媛、许鸿雁负责研究方案设计、实验操作、数据分析与论文撰写；董跃明、刘丹凤、孙鹏蕊负责数据采集与标注；颜锐、崔洪亮、雷红、任菲负责数据审核及内容审校。利益冲突：所有作者均声明不存在利益冲突 -
图 6 部分真菌原始显微镜图像和类激活热力图
A.光滑念珠菌；B.葡萄芽菌；C.新型隐球菌；D.黑曲霉菌；E.烟曲霉菌；F.杂色曲霉菌
(1)原始显微镜图像；(2)类激活热力图，颜色越趋于红色，表示该区域受网络结构的关注度越高表 1 改进的MobileNetV2真菌分类模型在测试集中的表现(x±s，%)
菌种 查准率 召回率 F1值 烟曲霉菌 99.77±0.16 98.95±0.29 99.36±0.08 杂色曲霉菌 96.41±0.62 99.77±0.16 98.06±0.25 黑曲霉菌 100±0 97.16±0.63 98.56±0.32 解脂假丝酵母菌 99.63±0 100±0 99.81±0 近平滑念珠菌 96.36±0.51 97.90±0.46 97.12±0.46 葡萄芽菌 97.50±0.70 96.53±0.17 97.01±0.40 季也蒙念珠菌 99.87±0.18 99.35±0.49 99.61±0.16 克柔念珠菌 100±0 100±0 100±0 光滑念珠菌 99.39±0.86 96.97±1.71 98.15±0.77 热带念珠菌 96.53±1.87 99.28±1.02 97.86±0.85 新型隐球菌 100±0 99.57±0 99.78±0 
下载: 导出CSV
表 2 不同分类模型在测试集中的运行结果
指标 参数量(M) 内存占用量(M) FPS 准确率(x±s，%) AUC(x±s) ResNet18 11.18 252.44 527 98.83±0.16 0.9935±0.0008 MobileNetV2 2.24 603.82 501 98.68±0.12 0.9928±0.0006 SENet 0.74 292.24 562 97.78±1.70 0.9878±0.0032 DenseNet121 6.97 1341.44 177 98.81±0.27 0.9939±0.0011 Inception V3 21.81 468.05 105 98.84±0.11 0.9948±0.0007 改进的MobileNetV2 4.22 356.89 573 99.09±0.18 0.9944±0.0018 FPS：网络每秒处理的图像数量；AUC：曲线下面积
下载: 导出CSV
-
[1] Chen M, Xu Y, Hong N, et al. Epidemiology of fungal infections in China[J]. Front Med, 2018, 12: 58-75. doi: 10.1007/s11684-017-0601-0 [2] Sanguinetti M, Posteraro B, Beigelman-Aubry C, et al. Diagnosis and treatment of invasive fungal infections: looking ahead[J]. J Antimicrob Chemother, 2019, 74: 27-37. [3] 何文军, 李曼, 李涛, 等. 基于血细胞形态识别的自动检测系统的研发[J]. 现代检验医学杂志, 2019, 34: 110-114. https://www.cnki.com.cn/Article/CJFDTOTAL-SXYN201902027.htm He WJ, Li M, Li T, et al. Study on Automatic Detection System Base on Blood Cell Morphology Recognition[J]. Xiandai Jianyan Yixue Zazhi, 2019, 34: 110-114. https://www.cnki.com.cn/Article/CJFDTOTAL-SXYN201902027.htm [4] 赵颖, 李志荣, 赵建宏, 等. 河北地区临床实验室丝状真菌分离鉴定情况分析[J]. 中国真菌学杂志, 2020, 15: 206-212. doi: 10.3969/j.issn.1673-3827.2020.04.004 Zhao Y, Li ZR, Zhao JH, et al. Analysis of filamentous fungi isolations from clinical laboratories in Hebei province[J]. Zhongguo Zhenjunxue Zazhi, 2020, 15: 206-212. doi: 10.3969/j.issn.1673-3827.2020.04.004 [5] Tamiev D, Furman PE, Reuel NF. Automated classification of bacterial cell sub-populations with convolutional neural networks[J]. PLoS One, 2020, 15: e0241200. doi: 10.1371/journal.pone.0241200 [6] Kulwa F, Li C, Zhang J, et al. A new pairwise deep learning feature for environmental microorganism image analysis[J]. Environ Sci Pollut Res Int, 2022. doi: 10.1007/s11356-022-18849-0. [7] Zhang J, Li C, Kosov S, et al. LCU-Net: A novel low-cost U-Net for environmental microorganism image segmentation[J]. Pattern Recogn, 2021, 115: 107885. doi: 10.1016/j.patcog.2021.107885 [8] Liang CM, Lai CC, Wang SH, et al. Environmental microorganism classification using optimized deep learning model[J]. Environ Sci Pollut Res Int, 2021, 28: 31920-31932. doi: 10.1007/s11356-021-13010-9 [9] 李卓识, 陈晓旭, 温长吉, 等. 基于机器学习的鹅膏属真菌形态特征分类模型研究[J]. 中国农机化学报, 2020, 41: 136-143. https://www.cnki.com.cn/Article/CJFDTOTAL-GLJH202001026.htm Li ZS, Chen XX, Wen CJ, et al. Study on classification model of morphological characteristics of amanita fungi based on machine learning[J]. Zhongguo Nongjihua Xuebao, 2020, 41: 136-143. https://www.cnki.com.cn/Article/CJFDTOTAL-GLJH202001026.htm [10] Tahir MW, Zaidi NA, Rao AA, et al. A fungus spores dataset and a convolutional neural network based approach for fungus detection[J]. IEEE Transact Nano Biosci, 2018, 17: 281-290. [11] Zhang J, Lu S, Wang X, et al. Automatic identification of fungi in microscopic leucorrhea images[J]. J Opt Soc Am A Opt Image Sci Vis, 2017, 34: 1484-1489. doi: 10.1364/JOSAA.34.001484 [12] 周院. 基于深度学习的真菌图像分类算法研究[D]. 西安: 西安理工大学, 2019. [13] 郝如茜. 白带显微图像中霉菌自动识别及清洁度判定的研究[D]. 成都: 电子科技大学, 2017. [14] Otsu N. A threshold selection method from gray-level histograms[J]. IEEE Transact Syst Man Cyb, 2007, 9: 62-66. [15] Hao R, Wang X, Zhang J, et al. Automatic detection of fungi in microscopic leucorrhea images based on convolu-tional neural network and morphological method[C]. 2019 IEEE 3rd Information Technology, Networking, Electronic and Automation Control Conference (ITNEC), 2019: 2491-2494. [16] Sandler M, Howard A, Zhu M, et al. Mobilenetv2: Inverted residuals and linear bottlenecks[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 4510-4520. [17] Deng J, Dong W, Socher R, et al. Imagenet: A large-scale hierarchical image database[C]. 2009 IEEE Conference on Computer Vision and Pattern Recognition, 2009: 248-255. [18] 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. [19] Yu F, Koltun V. Multi-Scale Context Aggregation by Dilated Convolutions[C]. International Conference on Learning Representations, 2016. https://doi.org/10.48550/arXiv.1511.07122. [20] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 770-778. [21] Huang G, Liu Z, Laurens V, et al. Densely connected convolutional networks[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 4700-4708. [22] Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the inception architecture for computer vision[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 2818-2826. [23] Loshchilov I, Hutter F. SGDR: Stochastic Gradient Descent with Restarts[J]. ICLR 2017 Conference Paper, 2016. https://doi.org/10.48550/arXiv.1608.03983. doi: 10.48550/arXiv.1608.03983 [24] Pan SJ, Qiang Y. A survey on transfer learning[J]. IEEE Transact Knowl Data En, 2010, 22: 1345-1359. [25] Selvaraju RR, Cogswell M, Das A, et al. Grad-CAM: Visual explanations from deep networks via gradient-based localization[J]. Int J Comput Vision, 2020, 128: 336-359. [26] Mital ME, Tobias RR, Villaruel H, et al. Transfer learning approach for the classification of conidial fungi (genus aspergillus) thru pre-trained deep learning models[C]. 2020 IEEE Region 10 Conference (Tencon), 2020: 1069-1074. [27] Billones RK, Calilung EJ, Dadios EP, et al. Aspergillus species fungi identification using microscopic scale images[C]. 2020 IEEE 12th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management(HNICEM), 2020. doi: 10.1109/HNICEM51456.2020.9400039. [28] Zawadzki P. Deep learning approach to the classification of selected fungi and bacteria[C]. 2020 IEEE 21st International Conference on Computational Problems of Electrical Engineering (CPEE), 2020: 1-4. [29] Zieliński B, Sroka-Oleksiak A, Rymarczyk D, et al. Deep learning approach to describe and classify fungi microscopic images[J]. PLoS One, 2020, 15: e0234806. -
点击查看大图
图(6) / 表(2)
计量
- 文章访问数: 111
- HTML全文浏览量: 27
- PDF下载量: 15
- 被引次数: 0
