From parasites to pixels: Bronchopulmonary nematodes and artificial intelligence in sheep lung health

This work investigated sheep lung health through parasitological and an artificial intelligence (AI)-based approach. The initial two chapters focus on how AI has continuously played its expanding role in the veterinary context ranging from diagnostic decision support to precision livestock farming. In the third chapter, the proven concept of edutainment to break the cycle of parasitic disease is highlighted along with potential AI integration. In the experimental section of the thesis, Chapter 4 examines and discusses the prevalence of bronchopulmonary nematodes (BPN), parasite burden, and co-infection dynamics in 467 sheep lungs. The overall prevalence of BPN infection was 44.3%, confirming the wide distribution of these parasites in the study population. Muellerius capillaris was the most prevalent species, detected in 36.8% of examined animals, and was associated with a mean parasite burden of 65.5 larvae per infected lung. Comparing with the past studies in Sardinia, the results suggests that BPN infections are persistent in Sardinian sheep farming. In chapter 5, convolutional neural network (CNN) was applied to the automated detection and classification of lung lesions using digital images. A diagnostic pipeline was developed to distinguish between healthy and pathological lungs, including inflammatory and neoplastic conditions. The highest performance value was achieved for the Neoplastic class (0.99), followed by the Healthy class (0.85) and the Undetermined class (0.83), demonstrating strong classification performance. These results led to the development of a pilot version of web-based application for automated classification of healthy and diseases lungs (link: https://lung-classification.streamlit.app/). In conclusion, this thesis demonstrates that bronchopulmonary nematodes are highly prevalent in sheep and are associated with significant lung pathology. Furthermore, it shows that AI-based image analysis can accurately and efficiently detect lung lesions, leading to improved respiratory disease surveillance, standardization of slaughterhouse inspections, and evidence-based sheep health management.