New Study Validates Travv's Cardiac Measurement Models
This month, a validation study of Travv’s cardiac measurement models was published in PLOS ONE. The article is open access, and we are sharing this because it reflects the kind of scientific standard we hold for putting models out in front of veterinarians. In an unregulated industry already full of companies releasing questionable AI applications, Travv’s mission is to stand on a pillar of scientific integrity.
Travv is committed to developing and commercializing AI that vets actually want, in the workflows they already use, based on transparent and rigorous validation.
Why does this matter clinically?
Myxomatous mitral valve disease (MMVD) is the most common heart disease in dogs, accounting for roughly 75% of all cardiac cases. The 2016 EPIC study evaluating 360 dogs with MMVD concluded that administration of pimobendan in affected dogs with cardiomegaly resulted in 60% prolongation of the preclinical period by approximately 15 months on average.
Measuring VHS and VLAS on canine thoracic radiographs is the gate for determining eligibility for pimobendan therapy in dogs with MMVD that don’t have access to echocardiography, per the pimobendan label and in accordance with ACVIM guidelines.
But getting the VHS and VLAS manually requires multiple measurements and calculations, and is subject to high interobserver variability. Automating this process with high-performing models not only creates clinical efficiencies and streamlines workflows, but provides accurate and repeatable measurements.
About Travv’s VHS/VLAS Models
Trained on over 11,000 radiographs, these models utilize convolutional neural networks to detect objects on radiographs and measure Vertebral Heart Size (VHS) and Vertebral Left Atrial Size (VLAS). Developed by Dr. John Craig at MetronMind, these models have been integrated into Travv PACS and workflows as part of the acquisition of MetronMind assets. To date they have analyzed more than 40,000 real-world images.
The Validation Study — PLOS ONE, 2026
The retrospective, single-center study evaluated automated measurement of two established radiographic indices of cardiac size: vertebral heart size (VHS) and vertebral left atrial size (VLAS).
Across 1,058 client-owned dogs spanning 80 breeds, the models produced measurements in close agreement with a trained observer, with a correlation of r = 0.947 for VHS and r = 0.811 for VLAS.
The mean differences from the human reference were 0.08 and 0.07 vertebrae, small enough that they would not be expected to change a clinical interpretation. On repeated reads of the same image, the models returned identical results every time. The study confirmed prior findings that right lateral thoracic radiograph views are preferred for measuring VHS and VLAS.
The takeaway? These models measure VHS and VLAS as reliably as a trained observer, and they do it the same way every time. Embedded directly into the imaging workflow, they take friction and time out of the veterinarian’s day without giving up human-level performance. These are the kinds of tools Travv is building across diagnostics, and this is just the beginning.
References
- Sykes KT, Gordon SG, Craig JJ, Wesselowski S, Watson AH. Validation of deep learning enabled software MetronMind to measure vertebral heart size and vertebral left atrial size in dogs. PLOS ONE. 2026. doi:10.1371/journal.pone.0339253.
- Boswood A, Häggström J, Gordon SG, et al. Effect of pimobendan in dogs with preclinical myxomatous mitral valve disease and cardiomegaly: the EPIC study—a randomized clinical trial. J Vet Intern Med. 2016;30(6):1765–1779.
- Keene BW, Atkins CE, Bonagura JD, et al. ACVIM consensus guidelines for the diagnosis and treatment of myxomatous mitral valve disease in dogs. J Vet Intern Med. 2019;33(3):1127–1140.
- Boehringer Ingelheim Animal Health USA Inc. VETMEDIN-CA1 (pimobendan) chewable tablets: prescribing information. Duluth, GA: Boehringer Ingelheim Animal Health USA Inc.