Detection of schistocytes is an important first step in the differential diagnosis of thrombotic microangiopathy. It is however labor intensive and prone to subjectivity. To improve and standardize the detection and quantification of schistocytes, we studied its automated analysis by digital microscopy DM1200 (CellaVision®) on 63 positive and 102 negative smears obtained from SP-50 (Sysmex®). Easy to use and very useful for staff training, it showed a lower between-observer coefficient of variation than usually described for manual counting (25% vs. 50%). Very sensitive (100%) in pre-classification, the detection of schistocytes was highly sensitive (98.4%) and specific (96.8%) after reclassification (AUCROC =0.9929), showing a good correlation with manual microscopy. TAT was comparable to manual counting. For positive smears, the percentage of schistocytes was similar between pre- and post-classification. However, 29.6% of pre-classified schistocytes were removed and 21.8% were added. For negative smears a significative overestimation of schistocytes (292%) was observed. Except poikilocytosis, on negative smears, the most common error of the software (24.9%) was due to platelets classified in schistocytes. Were also observed for example erroneous divisions of the image (3.2%) or artifactual schistocytes resulting from scratches in the smear (2.6%). Another limit is the high number of red blood cells not analyzed (46.8% for high-density smears), which might false the schistocytes percentage. To conclude, CellaVision® technology showed many benefits, but also limits that the operator needs to know.
Read full abstract