Deep learning automates vertebral heart score measurement on canine chest radiographs

A review of 17 studies finds deep learning, especially EfficientNet models, can automate canine vertebral heart score (VHS) estimation from thoracic radiographs, potentially reducing observer variability and aiding veterinary assessment of cardiac enlargement.

• Myxomatous mitral valve disease (MMVD) is the most common acquired cardiac disorder in dogs, and VHS is a standard radiographic metric for cardiac enlargement.
• Manual VHS measurement suffers from interobserver variability; automated methods using transfer learning have shown strong performance despite limited annotated veterinary datasets.
• Current limitations include differences in dataset annotation, a lack of multi-center validation, and the need for improved model interpretability before widespread clinical adoption.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.