Exploring the performance of an artificial intelligence- and morphology-driven workflow integrating 4 platelet enumeration technologies.

Objective: Platelet count, one of the main parameters of the complete blood count, requires accurate evaluation to guide patient management. It can be hampered by EDTA-induced pseudo-thrombocytopenia (PTCP), microcytic red blood cells (RBCs), RBC fragments, or giant platelets. A new set of 4 methods from the Mindray CAL-8000 platform, applicable on a single sample for platelet count, was evaluated.

Methods: The 4 options of the platform respectively use impedance (PLT-I); optical assessment (PLT-O) with a disaggregating agent; morphology (PLT-M) assessed by artificial intelligence-aided visualization on a smear prepared, stained, and analyzed by the platform; and PLT-Pro with morphologic assessment on a larger area of the smear. As part of an evaluation of the Mindray solution, a total of 2474 samples, collected on EDTA and sent for routine CBC, were further evaluated on the CAL-8000. The methods were combined according to a predefined algorithm.

Results: An automated report with accurate evaluation was ultimately obtained for 100% of the samples, using the sequence PLT-I, PLT-O, PLT-M, and PLT-Pro, which allowed accurate counting even in the presence of PTCP-related clumps.

Conclusions: Although this was a proof-of-concept assay including all analysis parameters, it validated the proposed new algorithm that can be implemented for routine flags.