In order to simplify the complexity of white blood cell classification in existing point-of-care testing (POCT) testing equipment, a white blood cell classification detection system based on microfluidic and multimode imaging was constructed. Microfluidic chip was used in the system. A multimodal optical imaging system based on the characteristics of blood samples was designed to obtain eigenvalue extraction of cells. Afterward, a BP neural network model was constructed to realize automatic classification of white blood cells. Finally, 80 human blood samples were classified and detected by this system and compared with the results of Sysmex XE-5000. The consistency correlation coefficients of white blood cells, lymphocytes, monocytes, neutrophils and eosinophils are 1.038, 0.907, 0.549, 0.922 and 1.028, respectively, and the CV values of the four types of white blood cells in the stability test were all below 10%. In this study, a white blood cell classification and detection system with small size, simple operation, fast single-sample detection, high accuracy, and no maintenance is required. It will provide a solid technical support for the further development of POCT blood cell analysis equipment.
Keywords: BP neural network; imaging systems; microfluidic chip; multimode imaging.
© 2020 Wiley-VCH GmbH.