Differential Diagnosis of Tuberculosis and Pneumonia in Cattle Lung Images Using Transfer Learning and YOLOv8: A Deep Learning Solution

Le résumé fourni par la source

Bovine tuberculosis (bTB) is a zoonotic, infectious, and chronic disease associated with significant economic losses, persisting as an endemic concern in various regions globally. Histopathological examinations are a crucial diagnostic method, particularly with the identification of Langhans giant cells, which serve as a distinctive marker. This study focuses on the differential diagnosis of tuberculosis from other types of pneumonia utilizing transfer learning. Additionally, it aims to enhance bTB diagnosis by employing the YOLOv8 algorithm to detect Langhans giant cells in histopathological images. While machine learning algorithms, such as logistic regression, SVM, and random forest, achieved a notable accuracy of 98.6% in the tuberculosis–pneumonia classification task, the YOLOv8 models demonstrated promising test-set performance in localizing Langhans giant cells, with mAP50 values ranging from 0.897 to 0.906. These findings offer a valuable contribution to the advancement of bTB diagnostic capabilities, particularly in distinguishing tuberculosis from other pulmonary conditions.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.