Utility of Faster R-CNN in methodological comparison and evaluation of reticulocytes

Abstract

Objective: The purpose of this study was to evaluate the methodological comparison of reticulocytes by using the intelligent learning system Faster R-CNN, a set of reticulocyte image detection systems developed using deep neural networks.

Methods: We selected 59 EDTA-K2 anticoagulated whole blood samples and calculated the RET% using seven different Sysmex XN full-automatic hematology analyzers with Faster R-CNN in the laboratory. We compared and evaluated the methods and statistically analyzed the correlation between the various test results.

Results: The results indicated a high degree of consistency between the seven Sysmex XN full-automatic hematology analyzers and Faster R-CNN in detecting RET%. The correlation coefficients were 0.987, 0.984, 0.986, 0.987, 0.987, 0.988, and 0.986, respectively.

Conclusion: We found that the Sysmex XN full-automatic hematology analyzers in our laboratory using the Faster R-CNN system met the requirements of the methodological comparison of reticulocyte detection and this intelligent learning system can be a useful clinical tool.

Keywords: clinical laboratory; hematology analyzers; intelligent learning system; methodological comparison; reticulocyte.