Objective To evaluate the clinical performance of an automated image analysis systems named CellaVision DM96 in classifying White Blood Cells. Methods A total of 2267 peripheral blood samples (male 1 235, female 1 034, average age 46) were obtained from outpatient and inpatient in Peking Union Medical College Hospital (PUMCH). These samples were selected to evaluate the precision, sensitivity, specificity and the analytical error of the system. We first evaluated the coincidence rate of reclassification and manual microscopy. On the base of favourable coincidence rate, we then evaluated the correlations between the pre–classification and reclassification of segmented neutrophil, band neutrophil, lymphocyte, monocyte, eosinophile, basophile, blast cell, promyelocyte, myelocyte, metamyelocyte, plasma cell and reactive lymphocyte. Results The sensitivity and specificity of pre–classification of White Blood Cell were 46%–100% and 24%–92%, respectively. When studied on the cell level, the total coincidence rate of pre–classification was 88%. And the coincidence rates of pre–classification and reclassification of White Blood Cell were 6%–95% and 25%–100%, respectively. When assessed on the sample level, the coincidence rates of pre–classification and reclassification of leukocytes were 64%–98% and 84%–100%, respectively. The correlations of pre–classification and reclassification of leukocytes in order from high to low were: lymphocyte, segmented neutrophil, eosinophile, band neutrophil, monocyte, basophile, when r were 0.943 9, 0.915 2, 0.785 4, 0.775 6, 0.676 2 and 0.289 1, respectively. The correlations between reclassification and manual microscopy of White Blood Cell were higher than those between pre–classification and manual microscopy. Order from high to low was: eosinophile, segmented neutrophil, lymphocyte, monocyte, band neutrophil, basophile. And r were 0.972 1, 0.968 5, 0.957 0, 0.831 9, 0.800 6 and 0.648 7, respectively. The ability of this automated image analysis systems at pre–classification in distinguishing between band cell and segment cell, atypical lymphocyte and normal lymphocyte was not good. Conclusion The performance of reclassification was better than pre–classification. The reclassification can be substitute for the microscopy inspection, and be used in the Clinical practice. (Chin J Lab Med, 2015, 38: 168–172) Key words: Leukocyte count; Cell shape; Automation, laboratory
Read full abstract