Abstract: Objective  To verify and evaluate the performance of CellaVision DM96 system (DM96).  Methods  A total of 422 peripheral blood samples were collected and prepared by automated smear and stain system， and then analyzed by DM96 white blood cell differentiation. The results of white blood cell differentiation by DM96 were compared with those by manual microscopy. The correlation of the results by DM96 and manual microscopy was analyzed by linear regression analysis. The coincidence rate of identifying cells based on original and enlarged reference cell libraries was calculated. The ability to identify important abnormalities through analyzing 110， 210，310 and 410 cells and the time cost  were calculated.  Results  The correlation coefficients of neutrophils and lymphocytes were 0.91 and 0.88 between DM96 and manual microscopy respectively， while the correlation coefficient of monocytes was only 0.31. DM96 agreed 27.56%，37.96%，41.85%，45.12% and 29.76% for pre-differentiation of promyelocytes， myelocytes， metamyelocytes ， blast cells and variation lymphocytes based on original reference cell library，and 33.23%，56.86%，48.33%，58.08% and 31.25% based on enlarged reference cell library. The ratios of identifying important abnormalities were 87.50%， 92.65%, 94.85% and 97.59% when analyzing 110， 210， 310 and 410 cells. It took (2.50±0.41) min per case by manual microscopy， (1.98±0.20) min per case by DM96 and (0.91±0.14) min per case by manual confirming. Time costs of analyzing based on original and enlarged reference cell libraries were (2.01±0.39) and (2.09±0.54) min per case. With the numbers of analyzed cells increasing， time costs increased simultaneously. Conclusions  The clinical application of DM96 could improve the automatization and standardization of white blood cell differentiation.

Key words: Peripheral blood cell analysis, Differentiation, Laboratory automation, Performance evaluation, Peripheral blood cell analysis, Differentiation, Laboratory automation, Performance evaluation