Comparison of ovine faecal Strongyle egg counts from an accredited laboratory and a rapid, on-site parasite diagnostic system utilising a smartphone app and machine learning

Abstract

Traditional treatment for gastrointestinal helminths in grazing livestock often involves untargeted, metaphylactic blanket treatment of animals with anthelmintics. As a result, resistance to anthelmintic drugs has become a significant issue for farmers and veterinarians worldwide, impacting farm profitability and animal welfare. Faecal egg counts (FECs) are an important diagnostic test to combat further anthelmintic resistance as they enable practitioners to better distinguish between animals that require treatment and those that do not. FECs are labour-intensive, time-consuming and require trained personnel to process the samples and visually identify the parasite eggs. Consequently, the time between sample collection, transport, analysis, results, and treatment can take days. This study aimed to evaluate a rapid, on-site parasite diagnostic system utilising a smartphone app and machine learning in terms of its capability to provide reliable egg counts while decreasing the turnaround time for results associated with outsourcing the analysis. A total of 105 ovine faecal samples were collected. Each sample was homogenised and split equally between two containers. One container per sample was processed using the on-site, app-based system, the second container was sent to an accredited laboratory. Strongyle egg counts were conducted via video footage of samples by the system’s machine learning (ML) and a trained technician (MT) and via microscopic examination by an independent laboratory technician (LAB). Results were statistically analysed using a generalised linear model using SAS® (Version 9.4) software. The ratio of means was used to determine non-inferiority of the ML results compared to the LAB results. Both system egg counts (ML and MT) were higher (p < 0.0001) compared to those obtained from the laboratory (LAB). There was no statistically significant difference between the ML and MT counts. The app-based system utilising machine learning has been found to be non-inferior to the accredited laboratory at quantifying Strongyle eggs in ovine faecal samples. With its quick result turnaround, low outlay cost and reusable components, this portable diagnostic system can help veterinarians to increase their testing capacity, perform on-farm testing and deliver faster and more targeted parasite treatment to combat anthelmintic resistance.