病理切片疫苗安全性指标自动检测技术  

Automatic detection of vaccine safety indicators of pathological sections

作　　者：宋泽园 穆光慧 胡美容 戴博[1] 张大伟[1] SONG Zeyuan;MU Guanghui;HU Meirong;DAI Bo;ZHANG Dawei(Engineering Research Center of Optical Instrument and System,Ministry of Education,Shanghai Key Laboratory of Modern Optical System,University of Shanghai for Science and Technology,Shanghai 200093,China;Guangdong Winsun Biopharmaceutical Co.,Ltd.,Guangzhou 511356,China)

机构地区：[1]教育部光学系统工程仪器与研究中心,上海市现代光学系统重点实验室,上海理工大学光电信息与计算机工程学院,上海200093 [2]广东永顺生物制药股份有限公司,广州511356

出　　处：《分析试验室》2023年第1期22-32,共11页Chinese Journal of Analysis Laboratory

基　　金：国家重点研发计划项目(2016YFD0500603);国家自然科学基金项目(61775140)资助。

摘　　要：针对猪病灭活疫苗研发、生产过程中和疫苗病理安全性检测时,存在对油乳佐剂残留、炎症细胞聚集2大主要安全性指标评估人工统计不准确、效率低的问题,建立了一套小型病理切片的病理图像采集显微系统,并提出了针对疫苗安全性指标的定量分析算法,实现了对病理切片快速数字化图像采集,以及对待测指标的定量统计分析及检测结果的可视化标记。搭建了一套高分辨率病理图像采集显微系统,并建立了炎症细胞聚集区域病理图像分割模型和油乳佐剂疫苗残留区域病理图像分割模型。对整张数字病理图像进行窗口扫描,得到整张病理图像2大指标的分析结果并进行可视化视觉标注。结果表明,此检测系统能在数秒内完成病理切片的病理图像采集,并对关键疫苗病理安全性检测指标进行可视化标记和定量分析,对油佐乳剂残留像素检出覆盖率误差为0.015,检测像素交并比为0.89,炎症细胞聚集平均像素检出覆盖率误差为0.01,检测像素交并比为0.74,实现了对病理安全性指标高精度无干预的自动检测分析和标记,可以满足实际生产和研究的检测需求。During the development of inactivated vaccines for swine diseases, the two main safety indicators, which are the proportion of oil emulsion adjuvant residues and the proportion of inflammatory cell aggregation in the vaccine pathological safety test, are assessed and manually counted by pathologists, resulting in subjective and inaccurate diagnosis results. At the same time, manual data collection has low efficiency and poor accuracy in the process of digitalization of animal vaccine research and development, and the problem of inability to use data rules to guide vaccine research and development needs to be solved urgently. Herein, a set of miniaturized microscopy system for collecting images of pathological slices and a quantitative analysis algorithm for vaccine safety indicators was established to realize the digital pathological image collection of pathological slices, the quantitative statistical analysis and result visualization. First of all, a high-resolution pathological image acquisition microscope system was established. Subsequently, the pathological image segmentation model of the inflammatory cell aggregation area and the pathological image segmentation model of the oil emulsion adjuvant vaccine residual area were established. Finally, window scanning was performed on the entire digital pathology image to obtain the two major labels and visual annotation results of the entire pathology image. Experiments results showed that this detection system can complete pathological image acquisition of pathological slices and visual labeling, and perform the quantitative analysis of key vaccine pathological safety detection indicators within a few seconds. The detection coverage rate error of oil emulsion adjuvant residue was 0.015, and the crossover ratio was 0.89. The detection coverage rate error of inflammatory cell aggregation was 0.01, and the crossover ratio was 0.74. It realized high-precision, non-interventional automatic detection, analysis and marking of pathological safety indicators, which can me

关 键 词：疫苗安全性检测 数字病理学 图像识别 

分 类 号：O657.31[理学—分析化学] X832[理学—化学]