A dedicated deep learning workflow for automatic Fasciola hepatica and Calicophoron daubneyi egg detection using the Kubic FLOTAC microscope.

Abstract

Fasciola hepatica and Calicophoron daubneyi are trematodes with significant health and economic impacts on ruminant livestock farms. An effective and reliable diagnosis is essential to control their spread. To improve copromicroscopic diagnosis, the Kubic FLOTAC Microscope (KFM), a portable digital microscope, was designed for both laboratory and field use. It is based on the use of FLOTAC/Mini-FLOTAC techniques and combines their high sensitivity, accuracy and precision with a reliable system based on an Artificial Intelligence (AI) predictive model. It features automated parasite egg detection, powered by an integrated battery, a web interface for microscope control, and a dedicated AI server for image analysis. In this study, the system was optimized to better discriminate between the eggs of these two parasites through additional processing steps and a robust detection model. Two protocols, egg-spiked samples and naturally infected samples, were used to simulate different sample conditions, creating a dataset for model training and evaluation. A second dataset of field samples, with egg counts verified by optical microscopy, was used to assess performance. The detection performance during the evaluation of samples from both protocols was found to be satisfactory. Specifically, the average fecal egg count, obtained through the clinical report generated by the KFM system, exhibited a mean absolute error of only 8 eggs per sample. This result demonstrates that the KFM is a valuable tool for parasitological diagnosis that supports the livestock industry.