A Novel Automated Hematopathologist-Styled Commenter for Complete Blood Count and Smear Review

Abstract Introduction/Objective Prevention of errors and increasing productivity are essential for managing the workflow in hematopathology labs, particularly under stress and fatigue conditions. The proposed automated commenter system is designed to directly address these critical issues. Care was also taken to ensure ease of use and minimum training time. Methods/Case Report Using 20 quantitative and qualitative parameters in 3 groups (RBCs, WBCs and platelets), our novel automated commenter is able to generate a hematopathologist’s report/interpretation for CBC and smear review. Built-in reference ranges can be adjusted to accommodate variability in cutoffs. There are alarms for entry of seemingly odd values, differential counts not adding up to 100% or suspected cold agglutination, lipemia and hemolysis in the specimen. Our automated commenter was developed and tested against more than 1,000 hematopathology reports. These reports included clinicopathologic correlations for different type hematopathology neoplasms and benign disorders. At least 100 reports were selected for each disease category covering all subclasses and grades. Ample user acceptance testing was conducted using feedback surveys from hematopathologists, hemato-oncologists and clinicians. Results (if a Case Study enter NA) Generated comments are narrative, concise, comprehensive and accurate. The abnormal findings are arranged by their grade and significance. For example, the following comment was generated using the system: “CBC and smear review reveals mild macrocytic pancytopenia with anisocytosis, increased peripheral circulating blasts above the cutoff of Acute Leukemia and critically severe neutropenia (Hgb 11.0 g/dL, MCV 103.5 fL, RDW 14.9%, WBC 3.2 K/µL, blasts 29%, promyelocytes 2%, myelocytes 5%, metamyelocytes 1%, nRBCs 5/100 WBCs, ANC 0.3 K/µL & Plt 125 K/µL) with mild dysgranulopoiesis and dacrocytosis (teardrops).” Conclusion Absence of clerical and scientific errors, reportedly positive user experience and saving 50-80% of pathologists’ working time are all advantages that indicate feasibility of large-scale adoption of our system by hematopathology labs. Workflow fitness, standardization, minimizing the manual variations and using the system as a professional training tool also strongly support this conjecture. The system could be used to fully automate clinicopathologic correlation with phenotypic and genetic tests and to develop an automated test selector (Triage Advis