Detecting Medial Patellar Luxation With Ensemble Deep Convolutional Neural Network Based On A Single Rear View Image Of The Hindlimb

This study proposes a deep-learning model that diagnoses canine patellar luxation (PL) using a single rear-view image of dogs without the need to visit veterinary clinics. The performance of the model was compared with the diagnosis of veterinarians and gradient-weighted class activation mapping (Grad-CAM) was applied to visualize the key areas with representative features that are utilized for the classification process. The study enrolled 30 private animal clinics to collect images of dogs from July 2021 to July 2022, and two professional veterinarians were enlisted to select one of the three pictures from each dog.