SPLTRAK Abstract Submission
María Rodríguez, Javier Laguna, Alexandra Casanova, Judit Julián, Angel Molina, Josep Lluis Bedini, Anna Merino
Hospital Clínic, Barcelona, Spain

Introduction: Automatic systems for the classification of blood cells have been implemented in hematology laboratories during the last years. They are able to perform an automatic pre-classification of individual images of peripheral blood (PB) cells from the smear. However, these systems are not able to recognize most of the abnormal cell types circulating in blood in hematological diseases. The objective of this work is to compare the automatic pre-classification of cells in Mindray MC-80 and CellaVision DM9600 analyzers using abnormal samples.
Methods: 91 samples were included and belonging to patients with the following diagnosis: 46 acute leukemia(AL), 16 myelodysplastic(MDS) or myeloproliferative syndrome(MPS) and 29 lymphoma. Smears were stained in May Grünwald-Giemsa. Cell images were obtained using the following image analysers: 1) MC80 and 2) DM9600 analysers. The post-classification of the blood cell images was done by the clinical pathologist. To compare the obtained results, Passing Bablok regression was used.
Results: Correlation between the blast cell pre-classification post-classification values by the expert showed higher values in MC80 than those obtained in DM9600 (r=0.987 and 0.915, respectively)(Figure1). For AL cases, none of the smears containing blast cells showed absence of them using the MC80(false negatives:0). Nevertheless, 1/46 samples containing blasts was classified as containing normal cells using the DM9600(2.17%). Whit respect to detection of abnormal promyelocytes, only available for the MC80, it was noted that in the 65% of AL cases some of the blasts were pre-classified as abnormal promyelocytes.  
Considering the MDS-MPD samples, the pre-classification values ​​for blast cells compared to that of the clinical pathologist showed a higher correlation using the MC-80(0.985) than with the DM9600(0.883). (Figure2) The pre-classification of abnormal lymphocytes is available only in MC80 and not in DM9600. In a total of 9 patients with lymphoma(31%), the MC80 detected abnormal lymphoid cells in the corresponding smear.
Conclusions: MC80 showed a good performance for blast detection in peripheral blood samples corresponding to AL or MDS/MPS patients. When comparing blast cell preclassification values ​​obtained by the MC80 and DM9600 and those of the post-classification performed by the clinical pathologist, a higher correlation was obtained using the MC80. MC80 detected the presence of abnormal lymphocytes in almost a third of lymphoma cases, while DM9600 in none.