Head of Parasitology, Zoetis Reference Laboratories Zoetis Parsippany, New Jersey, United States
Abstract:
Background: The VETSCAN IMAGYST AI Fecal application utilizes a deep learning artificial intelligence (AI) algorithm to detect parasite ova, cysts and oocysts in canine and feline feces. Deep learning AI algorithms evolve by learning from additional data sets provided by medical experts. Regular evaluation of algorithm performance and usability are essential to ensure the highest quality on-market device.
Hypothesis/
Objectives: The VETSCAN IMAGYST AI Fecal algorithm (v.8293) will correctly identify the diagnostic stages of Ancylostoma, Cystoisospora, Giardia, Toxocara and Trichuris in fecal samples of naturally infected dogs and cats previously scanned and manually evaluated by diagnostic parasitology experts from multiple geographies across the US.
Animals: 853 fecal samples from a mix of client-owned animals undergoing routine fecal examinations (502 canine, 203 feline, and 148 Giardia (dog and cat)).
Methods: In previous studies, the 853 fecal samples were processed, manually read by diagnostic parasitology experts and digitally stored for future evaluation. In this study, those digitally scanned images were analyzed using the VETSCAN IMAGYST AI Fecal algorithm (v.8293) and compared against the original diagnostic parasitology expert manual reads.
Results: Combined diagnostic sensitivity and specificity for canine and feline samples for Ancylostoma, Toxocara, Cystoisospora, and Giardia ranged from 90.7-95.5% and 96.0-98.8%, respectively. Sensitivity and specificity for canine Trichuris was 93.1% and 99.7%, respectively.
Conclusions and Clinical Importance: The results of the VETSCAN IMAGYST AI Fecal algorithm tested in this study were comparable to those of diagnostic parasitology experts.
