Study tests AI to identify common canine skin lesions from photos

Canine dermatology researchers have reported a new image-based AI approach that can identify four common lesion types in dogs, erythema, lichenification, alopecia, and erosion or ulceration, from clinical photographs with reported accuracies above 90%, according to a 2026 Veterinary Dermatology study by Soh-Yoon Kang and colleagues. The work focuses on lesion-level recognition rather than diagnosing a specific disease, an important distinction in dermatology where visible lesion patterns often guide the next diagnostic step. The alopecia model performed best among the four, suggesting some lesion classes may be more readily captured by convolutional neural network models than others. (eurekamag.com)

Why it matters: For veterinary teams, the study points to a potentially useful triage and documentation tool, especially in busy general practice, teledermatology intake, and follow-up monitoring where consistency can be hard to maintain across clinicians and image quality varies. Common canine skin diseases, including atopic dermatitis, often present with signs such as erythema, self-induced alopecia, excoriations, and, with chronicity, lichenification, so tools that standardize lesion recognition could support faster workups, clearer records, and more objective tracking over time. Still, this appears to be an early clinical-support application, not a replacement for cytology, history, physical exam, or dermatology referral. (pubmed.ncbi.nlm.nih.gov)

What to watch: The next question is whether these models can hold up in real-world practice across varied lighting, coat types, skin pigmentation, body sites, and smartphone-quality images, and whether they move from lesion detection into validated clinical decision support. (ovid.com)