摘要
目的建立血细胞形态自动检测系统工作平台。方法运用自动化控制、显微数字影像与人工智能神经网络技术,自动拍摄待检样品选定区域的显微镜放大图像,提取血细胞图像面积、边界和纹理等多种特征,对血细胞形态进行识别。结果该系统识别白细胞的正确率95.1%,重复率0.1%,漏检率0.7%,稳定性97.7%和速度≤5min/片,该系统可估算有核红细胞、网织红细胞数量,在此基础上给出白细胞个数修正因子,同时可初步估算血小板数量,提示异常红细胞。结论实验设计了一个完整的血细胞形态识别系统,减轻了临床的工作量,提高了检测效率。
Objective To establish a blood cell morphology automatic detection platform.Methods Using the techniques of automatic control,digital microscopic imaging and artificial intelligence neural network,automatic shotting the selected area on the microscope magnified image of sample,then the area,boundary,texture and other characteristics were extracted to compare with blood cell in cell morphology.Results In identifying WBC,the system accuracy was 95.1%,repetition rate was 0.1%,miss rate was 0.7%,the stability was 97.7%,and the read speed less the 5min/piece.The system could estimate reticulocyte and nucleated red blood cell number.And then leukocyte number correction factor was given too.At the same time,he system could preliminary estimate platelet number and suggest abnormal red blood cells.Conclusion A complete blood cell morphological identification system was established wich could reduce the work load and improve the detection efficiency.
作者
何文军
李曼
李涛
钟伟国
肖长周
钟彦晶
张云超
徐宁
HE Wen-jun;LI Man;LI Tao;ZHONG Wei-guo;XIAO Chang-zhou;ZHONG Yan-jing;ZHANG Yun-chao;XU Ning(the Second Affiliated Hospital of Guangzhou University of Traditional Chinese Medicine/ Guangdong Provincial Hospital of Traditional Chinese Medicine,Guangzhou 510370,China)
出处
《现代检验医学杂志》
CAS
2019年第2期104-108,共5页
Journal of Modern Laboratory Medicine
基金
广东省科技支撑计划资助项目(粤科规财字[2016]88号-25)
关键词
血细胞形态
自动化检测
系统评价
hemocyte morphology
automatic detection
systematic reviews
