Veterinary dermatology · 2026
Artificial Intelligence-Based Identification of Common Canine Skin Lesions From Clinical Images.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Kang Soh-Yoon, Kang Yeong-Hun, Kim Hyun-Jung, Huh Eun-A, Jeon Jong-Soo, Kim Hyun-Su, Kim Min-Sun, Tiwari Anisha, Hwang Cheol-Yong
Аннотация
BACKGROUND: Accurate evaluation of skin lesions is an essential component of dermatological examination, yet it can be time-consuming and subject to interobserver variability. While artificial intelligence (AI) models have shown reliability in diagnosing specific skin diseases, lesion-level identification remains underexplored in veterinary dermatology.
OBJECTIVES: To develop and evaluate a convolutional neural network (CNN)-based AI model for the automated identification of four skin lesion types in dogs: erythema, lichenification, alopecia and erosion/ulcer.
ANIMALS: Clinical skin images were collected from dogs presented to a veterinary medical teaching hospital.
MATERIALS AND METHODS: Four EfficientNet models were independently trained, one for each lesion type. Model performance was evaluated by comparing the prediction results with veterinary surgeon-labelled data, using six metrics: accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and F1 score.
RESULTS: All four CNN models achieved an accuracy of > 90%, indicating a reliable performance across all lesion types. The alopecia model yielded the highest accuracy (98.12%) and F1 score (98.18%). The erythema and erosion/ulcer models exhibited balanced performance across all metrics. For the lichenification model, the sensitivity (87.02%) and F1 score (89.88%) were the lowest among the four lesion types.
CONCLUSIONS AND CLINICAL RELEVANCE: The CNN-based AI models developed in this study demonstrated validity in identifying common canine skin lesions from clinical images. These models may facilitate rapid and objective dermatological evaluation, supporting clinical diagnosis and lesion monitoring throughout treatment.
OBJECTIVES: To develop and evaluate a convolutional neural network (CNN)-based AI model for the automated identification of four skin lesion types in dogs: erythema, lichenification, alopecia and erosion/ulcer.
ANIMALS: Clinical skin images were collected from dogs presented to a veterinary medical teaching hospital.
MATERIALS AND METHODS: Four EfficientNet models were independently trained, one for each lesion type. Model performance was evaluated by comparing the prediction results with veterinary surgeon-labelled data, using six metrics: accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and F1 score.
RESULTS: All four CNN models achieved an accuracy of > 90%, indicating a reliable performance across all lesion types. The alopecia model yielded the highest accuracy (98.12%) and F1 score (98.18%). The erythema and erosion/ulcer models exhibited balanced performance across all metrics. For the lichenification model, the sensitivity (87.02%) and F1 score (89.88%) were the lowest among the four lesion types.
CONCLUSIONS AND CLINICAL RELEVANCE: The CNN-based AI models developed in this study demonstrated validity in identifying common canine skin lesions from clinical images. These models may facilitate rapid and objective dermatological evaluation, supporting clinical diagnosis and lesion monitoring throughout treatment.