Canine Dermal Analyser: Harnessing Artificial Intelligence and Deep Learning to Revolutionize Canine Skin Disease Detection

Abstract

Skin diseases in dogs are a prevalent concern, requiring timely and accurate diagnosis for effective treatment. This study proposes an intelligent system that integrates a Convolutional Neural Network (CNN) with lesion segmentation and a weighted scoring algorithm to improve classification accuracy in multiple disease categories such as ringworm, mange, and yeast infection. Unlike traditional deep learning approaches, our method incorporates lesion segmentation to isolate affected areas prior to feature extraction, improving the model focus on disease-specific patterns. Additionally, a domain knowledge based weighted scoring algorithm refines predictions by combining CNN-derived probabilities with owner-provided symptom assessments. The methodology involves pre-processing images for noise reduction, lesion segmentation for targeted analysis, CNNbased feature extraction, and a scoring mechanism that weighs expert-defined symptoms. Experimental results demonstrate a classification accuracy of 98%, significantly enhancing reliability compared to conventional CNN-only models. This hybrid approach offers an efficient, cost-effective and mobile-friendly diagnostic tool, empowering both pet owners and veterinarians with rapid and precise skin disease identification, ultimately improving canine care.
Index Terms—Canine skin disease detection, convolutional neural networks (CNNs), image segmentation, veterinary diagnostics, machine learning, lesion detection, computer vision, deep learning, automated disease classification, weighted scoring algorithm