Artificial Intelligence-Based Identification of Common Canine Skin Lesions From Clinical Images.

AI-based CNN models reliably identify common canine skin lesions, supporting objective dermatological evaluation.

This study developed and validated convolutional neural network (CNN) models using EfficientNet architecture to automatically identify four common canine skin lesions: erythema, lichenification, alopecia, and erosion/ulcer from clinical images. Four independent models were trained on clinical images collected from a veterinary teaching hospital and evaluated against veterinary surgeon-labeled reference data using accuracy, sensitivity, specificity, PPV, NPV, and F1 score metrics. All models achieved >90% accuracy, with the alopecia model performing optimally (98.12% accuracy, 98.18% F1 score). The erythema and erosion/ulcer models demonstrated balanced performance across metrics, while the lichenification model showed relatively lower sensitivity (87.02%) and F1 score (89.88%). The findings support the clinical utility of AI-based systems for objective, rapid dermatological lesion identification, potentially reducing interobserver variability and supporting clinical diagnosis and treatment monitoring in canine dermatology.

This summary was distilled by AI and may occasionally misinterpret data. Confirm critical details with the primary literature before clinical application.