Inteligencia artificial en ecocardiografía canina: revisión narrativa de métodos, validación y retos técnicos

Objective: This review critically examined artificial intelligence (AI) applications in canine echocardiography, automated measurement and disease classification strategies, validation designs, and the requirements for safe clinical translation. Method: A narrative review was conducted. PubMed/MEDLINE was searched on 20 October 2026, supplemented by screening the reference lists of identified articles, and studies that applied machine learning or deep learning to canine echocardiograms were selected by relevance. Results: In the clinical veterinary setting, two studies addressed automated ventricular measurements through anatomical landmark localization and staging of myxomatous mitral valve disease through echocardiography–radiography fusion. An additional group of experimental studies applied deep learning to left ventricular segmentation and motion tracking in dogs. The main challenges identified were operator variability, breed-related anatomical heterogeneity, the lack of multicenter validation, and reliance on imperfect reference standards. Conclusions: The clinical usefulness of these models requires patient- and center-level evaluation, uncertainty quantification, agreement studies against specialists, and analysis of errors near treatment thresholds. A two-stage research pathway, measurement followed by classification, was proposed, complemented by prospective assessment and veterinary oversight.