Automated vertebral heart size estimation from thoracic radiographs in dogs with AI-assisted clinical decision support.

Cardiomegaly is a clinically significant indicator of cardiac disease in canine companion animals, where early and accurate detection is essential for effective treatment planning. Vertebral Heart Size (VHS) measurement from thoracic radiographs is a widely used method for assessing cardiac enlargement; however, manual measurement and report drafting can be time-consuming and subject to inter-observer variability. This study proposes an AI-assisted, integrated framework to support veterinarians in radiographic assessment of cardiac size by automating VHS estimation from canine thoracic DICOM radiographs. A deep learning-based computer vision model is trained to detect anatomical landmarks and compute VHS using a curated dataset of annotated images. The extracted VHS measurements, together with relevant DICOM metadata and structured clinical inputs, are subsequently processed by a Large Language Model (LLM) to generate a structured veterinary summary for clinician review. The proposed framework enables rapid VHS computation and automatic generation of a preliminary clinical summary within seconds of image upload. Quantitative evaluation of our deep learning pipeline demonstrates accurate detection of thoracic anatomical landmarks and reliable VHS estimation. In contrast, qualitative evaluations indicate that the generated veterinary summaries are coherent, context-aware, and consistent with radiographic findings. Collectively, these results demonstrate the potential of multimodal AI systems to enhance veterinary radiology workflows and serve as effective clinical decision-support tools for the timely assessment of cardiomegaly in companion animals.