Abstract
Gastrointestinal parasites are a major health and economic concern in small ruminants. The classic microscopic approach using the manual McMaster method serves to quantitatively count parasite eggs, which are labor-intensive and prone to variation. Artificial intelligence-based systems (Parasight(®), powered by Fecalsight AI™) could provide quicker and more objective alternatives; therefore, independent validation is necessary before clinical implementation. The objective of this study was to evaluate the agreement, classification consistency, and diagnostic performance of Parasight(®) relative to the manual McMaster method, with a focus on its suitability as a screening and decision-support tool. Fecal samples from 44 Kiko goats over 3 sampling times were analyzed using both methods, with manual counts performed independently by 2 observers. Agreement between methods was assessed using Lin's concordance correlation coefficient, Bland-Altman analysis, and Cohen's Kappa for categorical classification. Diagnostic performance for identifying animals exceeding the clinical treatment threshold (>1000 eggs per gram) was evaluated using receiver operating characteristic (ROC) analysis, and regression modeling was used to characterize associations between methods. Manual observers showed high reliability, confirming the suitability of the McMaster method as a reference. Compared with manual counts, Parasight(®) consistently underestimated egg counts, resulting in poor-to-moderate absolute agreement; however, it reliably ranked animals by parasite burden and showed excellent discrimination for identifying animals above the treatment threshold (AUC = 0.90-0.96). Regression analyses further demonstrated linear or curvilinear associations depending on egg counts. Overall, the Parasight(®) device reliably captured relative parasite burden but required a lower operational threshold to match manual treatment decisions.