Development of a Predictive Artificial Intelligence Model for Preoperative Fracture Severity Classification Using Radiographic Data from Nigerian Veterinary Hospitals

Accurate preoperative fracture severity classification is essential for selecting appropriate fixation strategies and forecasting surgical outcomes in veterinary orthopedics. The development of artificial intelligence-assisted radiographic classification tools in the Nigerian veterinary context presents an original opportunity to improve diagnostic consistency and surgical planning efficiency. This doctoral study develops, trains, and validates a deep convolutional neural network model for preoperative fracture severity classification using radiographic data from veterinary hospitals across all six Nigerian geopolitical zones between 2020 and 2024. An annotated dataset of 3,840 digital radiographic images of long bone fractures in dogs and cattle was compiled, with expert veterinary radiologist annotation using the AO Veterinary Fracture Classification system as the reference standard. A transfer learning approach using ResNet-50 architecture was employed, with training on 80% of the dataset and external validation on the remaining 20%. Model performance was evaluated by area under the receiver operating characteristic curve, sensitivity, specificity, and calibration. The developed model achieved an area under the curve of 0.91 for binary fracture type classification and 0.84 for multi-class severity grading. Inter-model agreement with senior radiologist classification was higher than junior clinician agreement in validation testing. The model demonstrated robust performance across image quality variations representative of Nigerian veterinary radiographic infrastructure. This doctoral contribution represents the first AI-assisted fracture classification tool developed specifically for Nigerian veterinary radiographic data, advancing the application of computational imaging in lower-resource veterinary surgical settings. Keywords: artificial intelligence, fracture classification, radiographic image analysis, convolutional neural network, veterinary radiology.