Predicting stage B2 myxomatous mitral valve disease in dogs using machine learning and routine clinical data.

Early identification of Stage B2 myxomatous mitral valve disease in dogs is critical for initiating appropriate treatment to delay the onset of heart failure. Although echocardiography is the gold standard for diagnosing Stage B2 cardiac remodeling, it has limitations in general clinical settings owing to the limited availability of specialized equipment and the need for advanced technical expertise. We developed a machine learning model using routine clinical data to identify dogs with Stage B2 disease without echocardiography. A total of 387 client-owned dogs (252 non-Stage B2 dogs and 135 Stage B2 dogs) were evaluated. A gradient boosting algorithm was trained on 80% of the data and validated on a test dataset comprising the remaining 20%, incorporating demographic, hematological, serum biochemical, urinalysis, and thoracic radiographic variables. The model achieved an accuracy of 0.825 [95% confidence interval (CI): 0.738-0.900], a sensitivity of 0.815 (95% CI: 0.650-0.957), a specificity of 0.830 (95% CI: 0.729-0.919), and an area under the receiver operating characteristic curve of 0.922 (95% CI: 0.860-0.971). These findings demonstrate that machine learning can accurately classify Stage B2 status using routine clinical data, thereby assisting veterinary clinicians in the early detection and specialist referral when echocardiography is unavailable.