(HM13) The Evaluation of the VETSCAN IMAGYST™ Artificial Intelligence (AI) Blood Smear Polychromatophil Determination

Abstract:

Background:
Peripheral blood film (PBF) analysis is a vital part of a comprehensive complete blood count (CBC). The VETSCAN IMAGYST AI Blood Smear Application (VS-I) uses a deep learning artificial intelligence algorithm to evaluate blood smears and provide an estimated white blood cell count and differential, estimated platelet count, and estimated polychromatophil count for canine and feline patients.

Hypothesis/

Objectives:
The VETSCAN IMAGYST AI Blood Smear Application will correctly classify and quantify polychromatophils on blood smears, and agreement with reticulocyte counts will be appreciated in paired sample slides evaluated by American College of Veterinary Pathology (ACVP) board-certified clinical pathologists (CP).

Animals:
No animals were utilized in this study. Canine and feline EDTA blood samples submitted to Zoetis Reference Laboratories were used to create blood smears.

Methods:
An incomplete block design randomly assigned PBF samples to CPs. Slides were created from each EDTA blood sample and stained by a quick-Romanowsky type stain (Wright-Geimsa/Diff-Quik) and standard new methylene blue (NMB). CP-reviewed PBF evaluations for reticulocytes using standard NMB served as gold standard, and were combined statistically to provide the basis for results comparison to the VETSCAN IMAGYST AI Blood Smear Application.

Results:
In slides stained with Wright-Giemsa, 100% of the VS-I evaluated slides where polychromatophils were counted were within the 95% prediction interval as calculated from the reticulocyte percentage from NMB stained slides.

Conclusions and Clinical Importance:
The VETSCAN IMAGYST AI Blood Smear Application demonstrated accurate classification and quantification of polychromatophils vs. reticulocyte counts, comparable to CP-confirmation of regeneration and WBC count.