[Leukocyte differentiation by MICRO21 system using neural networking as a type of artificial intelligence]

NeuralVision is a type of artificial intelligence that has the ability to identify specific microscopic objects. The MICRO21 system uses neural networking to analyze the features of blood cells; color, shape, size, and density, just as a technologist would. The basic nucleated blood cell processing procedure of the MICRO21 is divided into three general categories; location, segmentation, and classification. Classification takes place using the Neural Network and is fine tuned during post-processing. We evaluated the performance of the MICRO21 automated differential vs. manual differential method. The intra-assay of leukocyte differentiation by MICRO21 was fairly good in normal samples with normal leukocyte counts. The neutrophils, lymphocytes and eosinophils count correlated well between MICRO21 and visual count. Post-processing was performed after the Neural Network has initially classified the cell, in which the confidence levels of MICRO21 to visual count was 87%. By post-processing, coincidence ratio of cell-by-cell by MICRO21 and visual count has been improved to 97%. Based on the above, MICRO21 is an effective and efficient system for locating, storing and displaying images of white blood cells, in which the technologist could determine specific characteristics of cells in each laboratory. Furthermore, the system may be applied to any blood cell and would be important from the perspective of cytological and morphological analysis, such as blood cells in bone marrow and body fluid, in the future.