Head of Vetscan Imagyst Platform Zoetis Global Diagnostics - Medical Affairs Parsippany, NJ, United States
Abstract:
Background: Urine sediment exam is vital to a complete urinalysis. A deep learning AI tool for in-clinic, rapid, and consistent evaluation of urine sediment samples is currently lacking. Hypothesis/
Objectives: Vetscan Imagyst® (VSI) AI Urine Sediment will accurately identify red blood cells, white blood cells, struvite crystals, calcium oxalate dihydrate crystals, and rod and cocci bacteria in agreement with digital review by ACVP-boarded clinical pathologists (ACVP-CPs). Animals: No animals were used in this study. Urine samples included a mix of client-owned dogs and cats undergoing urinalysis for any reason submitted to Zoetis Reference Laboratories. Some samples were artificially created by spiking donor urine with necessary elements (e.g., RBC from whole blood). A total of 175 urine samples consisting of 98 canine (56%) and 77 feline (44%) were evaluated.
Methods: Samples were prepped via the VSI prep method, scanned, and evaluated digitally by 2 randomized, blinded ACVP-CPs who recorded results for urine elements as an average of 10, 40X fields. Algorithm performance was calculated as compared to ACVP-CP consensus, which served as gold standard.
Results: The VSI AI Urine Sediment algorithm reliably identified urine sediment elements. Sensitivity and specificity for urine sediment object classes ranged from 73-98% and 76-99%, respectively (Table 1). Combined bacteria identification had a PPV of 91%.Conclusions and Clinical Importance: VSI AI Urine Sediment algorithm tested in this study was comparable to ACVP-CPs in the identification of elements in urine sediment. In-clinic utilization of VSI AI Urine Sediment can provide a diagnostic tool for urine sediment evaluation.
