Harnessing AI to make expert-level ear diagnostics widely accessible

In the laboratory (NAVDLab) of Neoklis Apolstolopoulos, DVM, DECVD, EBVS, Artificial intelligence (AI), is utilized to develop a proof-of-concept model capable of distinguishing healthy canine ear canals, ear masses, and otitis from video-otoscopic images, offering a promising way to close a diagnostic gap.

Otitis externa—ear canal inflammation—is one of the most common reasons dogs visit a veterinarian. It causes discomfort, pain, and, if untreated, can lead to hearing loss. Accurate diagnosis typically requires an otoscopic examination, yet surveys show that nearly half of cases are initially misdiagnosed in general practice compared with specialist evaluation by veterinary dermatologists.

Already widely used in human medicine, AI can be successfully adapted for veterinary applications. Within AI, the field of computer vision focuses on extracting meaningful information from images and videos to make informed recommendations. It can analyze veterinary images and identify patterns that a trained specialist would recognize.

Until recently, no such models existed in small animal dermatology. Unlike many AI projects in medicine, NAVDLab not only develops applied AI models but also thoroughly tests and validates them to ensure reliability, clinical relevance, and real-world performance before deployment. This rigorous evaluation process addresses a common gap in both human and veterinary AI research, where models are often published without sufficient validation in diverse clinical settings.

Building on this foundation, the team is now creating an expanded AI-powered diagnostic tool for canine ear disease. The new model will classify ear canals into one of five clinically relevant categories. Crucially, the dataset is curated by a board-certified veterinary dermatologist, ensuring gold-standard annotations for every image. This expert oversight is essential for producing a clinically relevant and reliable model.

Once trained, the AI model could serve as a diagnostic support tool for veterinarians and veterinary nurses worldwide, including those working in shelters or underserved areas. A practitioner could use a digital otoscope to capture an image and receive an instant AI-generated assessment. This capability would enable earlier diagnoses, improve evaluation consistency, and reduce missed cases.

The model will also serve as a powerful educational resource for training veterinarians, technicians, and students in otoscopy. By comparing their own assessments with AI output, users will receive immediate feedback, reinforcing learning in both in-person and remote training environments.

The project aims to make expert-level diagnostic interpretation more widely accessible. By bridging the gap between primary care and specialist evaluation, this tool has the potential to improve animal welfare, reduce disparities in care, and support continuing professional development in veterinary medicine.

This work represents a novel intersection of veterinary dermatology, otology, and artificial intelligence. The vision is for an AI-augmented diagnostic ecosystem that complements—not replaces—veterinary clinical judgment, enhances training for students and practitioners, and supports research into the epidemiology and management of ear disease in dogs. 

About LSU Vet Med: Bettering lives through education, public service, and discovery

The LSU School of Veterinary Medicine is one of only 33 veterinary schools in the U.S. and the only one in Louisiana. LSU Vet Med is dedicated to improving and protecting the lives of animals and people through superior education, transformational research, and compassionate care. We teach. We heal. We discover. We protect.