Comparative evaluation of an AI-based counting system (OvaCyte™) and the McMaster counting method for quantification of strongyle eggs in sheep faeces

Abstract

We compared an artificial intelligence (AI)-based technology (OvaCyte™, OC) for the enumeration of strongyle eggs in sheep faeces with the McMaster method (MM). Initially, two experiments were performed with faeces containing pure Haemonchus contortus eggs. In experiment A, faeces containing three egg concentrations were processed using OC (extended and standard mode) in parallel with MM. In experiment B, faeces were spiked with different amounts of eggs. Secondly, samples from naturally infected sheep were analysed. Overall, EPG values in experiment A were consistent across all replicates at each dilution. Accuracy was particularly good for the AI-method (mean OC=72 %, mean MM=45 %), and it also achieved the highest precision (CV 5.6–40 %). In experiment B, as in experiment A, within replicate variability was observed at for both methods all concentrations. Although there were no significant differences between sample means, precision and the number of egg-positive samples was higher for OC. Finally, analysis of both experimental (r = 0.98) and field samples (r = 0.93) showed a strong positive correlation between OC and MM. OC also yielded a higher proportion of positive samples than MM in the field study OC provided a higher proportion of positive samples than MM. This study is the first comparison of OC and MM using both experimental and field-based data. In contrast to previous studies, our analysis was based on identical sample preparations that were processed in parallel using both methods. Although the results show strong agreement between methods, some limitations of OC were noted. These limitations can probably be overcome by further refinement of the AI model.