A reproducible differential blood cell count without a cytometer or specific markers

Abstract

Introduction: A differential blood cell count represents a substantial component of the daily clinical routine. The aim of this study was to conduct automated high-throughput analyses of light microscopy images of leukocytes with unspecific staining (hematoxylin-eosin, HE) and to classify the resulting subgroups (lymphocytes, monocytes and neutrophils). Materials and Method: The software package CellProfiler was used for the image analysis and cell classification. The results were compared to those from hematological laboratories that analyzed the same samples. Results: There was a high similarity between the results obtained by image analysis and those from the hematological laboratories, with an r²=0.95 for the differential count and an r²=0.89 for the total count. Conclusion: Differential blood cell count using light microscopy, nonspecific dyes and advanced image analysis is accurate, reproducible, and less expensive compared to other techniques. The results also showed a strong correlation when compared to traditional methods.

Keywords

cytometry - blood cell count - high-throughput analyses - image analyses - machine learning