Abstract: Objective　To establish an artificial intelligence (AI)⁃assisted platform for detection of parasite eggs, and to evaluate its detection efficiency and accuracy, so as to provide technical supports for elimination of parasitic diseases.  Methods　A total of 1 003 slides of Enterobius vermicularis, horkworm, Trichuris trichiura, Clonorchis sinensis, Taenia, Ascaris lumbricoides, Schistosoma japonicum, Paragonimus westermani and Fasciolopsis buski eggs were collected, and converted into digital images with an automatated scanning microscope to create a dataset. Based on the Object Detection platform on the Baidu Easy DL model, an AI⁃assisted platform for detection of parasite eggs was created through procedures of uploading, labeling, training, evaluation and optimization. Then, 70% of the datasets were randomly selected for model training, and the precision, recall and average accuracy were calculated to evaluate the effectiveness of platform for recognition of parasite eggs. In addition, the platform was deployed on the computer and smart phone terminals for use. Results　An AI⁃assisted platform for detection of parasite eggs was successfully created. If the platform was deployed using the public cloud application programming interface (API), the average accuracy, precision and recall of the platform were 93.42%, 92.55% and 89.32% for recognition of parasite eggs. If the platform was deployed using the offline software development kit (SDK), the average accuracy, precision and recall of the platform were 92.97%, 94.78% and 87.63% for recognition of parasite eggs. In addition, the precision of the platform was 97.00% and 96.23% for identification of Taenia and C. sinensis eggs, respectively. Conclusions　The AI⁃assisted platform for detection of parasite eggs has been successfully created, which is high in the accuracy for recognition of parasite eggs and convenient in use. This platform may provide a powerful technical support for parasitic disease diagnosis. 

Key words: Parasite, Egg, Artificial intelligence, Detection efficiency

摘要： 目的　建立一种人工智能（artificial intelligence，AI）辅助寄生虫虫卵检测平台，评价其检测效率及精准度，为防控寄生虫病提供技术支持。方法　收集蠕形住肠线虫、钩虫、毛首鞭形线虫、华支睾吸虫、带绦虫、似蚓蛔线虫、日本血吸虫、卫氏并殖吸虫和布氏姜片吸虫等9种寄生虫虫卵玻片标本共1 003张，通过自动扫描显微镜将寄生虫虫卵玻片转化为数字图像，分类、整理并建立数据集。基于百度Easy DL“物体检测”平台，通过上传、标记、训练、评估优化等环节，建立基于AI的寄生虫虫卵检测平台。随机抽取70%的数据用于模型训练，计算精确率、召回率、平均精度以评价模型识别效果。将模型部署于电脑端及智能手机端，供用户使用。结果　成功建立了AI辅助的寄生虫虫卵检测平台。采用公有云应用程序接口（application programming interface，API）部署时，模型识别寄生虫虫卵的平均精度、精确率和召回率分别为93.42%、92.55%和89.32%；采用离线软件开发工具包（software development kit，SDK）部署时，模型识别寄生虫虫卵的平均精度、精确率、召回率分别为92.97%、94.78%、87.63%。该检测平台对带绦虫和华支睾吸虫虫卵的识别精确率分别为97.00%和96.23%。结论　成功建立了基于AI的寄生虫虫卵检测平台，其识别寄生虫虫卵精确率高、使用灵活，有望为寄生虫病诊断提供技术支持。 

关键词: 寄生虫, 虫卵, 人工智能, 检测效能