Comparison of the Diagnostic Accuracy of Semiautomated and Expert-Based Calculation of the Vertebral Heart Score in Cats and Dogs: First Stage Validation of the “Desktop Application for Automation of the VHS Calculation Based on Radiographs”

Introduction. Cardiac diseases are considered to be among the main causes of mortality in small domestic animals. Calculation of the vertebral heart score (VHS) is a method used to assess the cardiac silhouette enlargement on radiographs and predict the development of congestive heart failure. However, traditional manual procedure of VHS calculation is characterized by high inter-observer variability, subjectivity, and significant time-consumption. To improve the access to objective cardiac screening in every-day clinical practice, the authors has developed a desktop application for automating VHS calculation using the artificial intelligence (AI) technologies that unifies the assessment standard, minimizes human error, and accelerates processing of radiographs. The aim of the present pilot research is to conduct the first stage validation of the application and compare the diagnostic accuracy of two VHS calculation methods — semiautomated and expert-based.
Materials and Methods. A balanced sample of 30 right lateral chest radiographs of dogs and cats was selected for analysis as a result of simple random sampling. Preliminary VHS was calculated by manual measurements on the radiographs by the Buchanan & Bücheler method: two veterinarians specializing in visual diagnostics measured the short and long axes of the heart on the radiographs and then compared their dimensions against the spinal vertebrae. The semi-automated calculation of VHS was performed by the author-developed “Desktop application for automation of the vertebral heart score (VHS) calculation based on radiographs” using a convolutional neural network. App validation implied a comparison of the results of the two methods of VHS calculation, taking into account the intraclass correlation coefficient (ICC).
Results. The “Desktop application for automation of the vertebral heart score (VHS) calculation based on radiographs” has proved to perform a highly accurate and replicable automated calculation of VHS on lateral chest radiographs of dogs and cats. It also has demonstrated statistically high agreement with the calculations obtained by veterinarians specializing in visual diagnostics (ICC≥0.97), along with less time required for calculation.
Discussion and Conclusion. The first stage of app validation has demonstrated an excellent agreement between the AI and two specialists’ results, therefore, the proposed approach can be deemed a future-oriented tool for general practitioners. The present pilot research has several limitations, i.e. it was conducted on a relatively small sample of animals, which may be insufficient for evaluation of app performance in narrow subgroups (e.g., in specific breeds or in severe cardiomegaly cases). Further studies in larger, more heterogeneous cohorts, including patients with severe cardiomegaly and anatomical abnormalities, are required to confirm clinical significance of the results.

1. Ward JM, Youssef SA, Treuting PM. Why Animals Die: An Introduction to the Pathology of Aging. Veterinary Pathology. 2016;53(2):229–232. https://doi.org/10.1177/0300985815612151

2. Baisan A, Ciocan M, Bîrsan O, Vulpe V. Objective Assessment of the Radiographic Cardiac Size in Dogs – A Review of the Heart Size Measurements. Lucrari Stiintifice - Universitatea de Stiinte Agricole a Banatului Timisoara, Medicina Veterinara. 2017;50(2):16–23.

3. Buchanan JW, Bücheler J. Vertebral Scale System to Measure Canine Heart Size in Radiographs. Journal of the American Veterinary Medical Association. 1995;206(2):194–199.

4. Poad MH, Manzi TJ, Oyama MA, Gelzer AR. Utility of Radiographic Measurements to Predict Echocardiographic Left Heart Enlargement in Dogs with Preclinical Myxomatous Mitral Valve Disease. Journal of Veterinary Internal Med-icine. 2020;34(5):1728–1733. https://doi.org/10.1111/jvim.15854

5. Keene BW, Atkins CE, Bonagura JD, Fox PR, Häggström J, Fuentes VL, et al. ACVIM Consensus Guidelines for the Diagnosis and Treatment of Myxomatous Mitral Valve Disease in Dogs. Journal of Veterinary Internal Medicine. 2019;33(3):1127–1140. https://doi.org/10.1111/jvim.15488

6. Lord P, Hansson K, Kvart C, Häggström J. Rate of Change of Heart Size before Congestive Heart Failure in Dogs with Mitral Regurgitation. Journal of Small Animal Practice. 2010;51(4):210–218. https://doi.org/10.1111/j.1748–5827.2010.00910.x

7. Boissady E, de La Comble A, Zhu X, Hespel AM. Artificial Intelligence Evaluating Primary Thoracic Lesions Has an Overall Lower Error Rate Compared to Veterinarians or Veterinarians in Conjunction with the Artificial Intelligence. Veterinary Radiology and Ultrasound. 2020;61(6):619–627. https://doi.org/10.1111/vru.12912

8. Koo TK, Li MY. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Re-search. Journal of Chiropractic Medicine. 2016;15(2):155–163. https://doi.org/10.1016/j.jcm.2016.02.012. Erratum in: Journal of Chiropractic Medicine. 2017;16(4):346. https://doi.org/10.1016/j.jcm.2017.10.001

9. Sun Y, Wang X, Tang X. Deep Convolutional Network Cascade for Facial Point Detection. In: Proceedings of 2013 IEEE Conference on Computer Vision and Pattern Recognition. Portland, OR, USA: IEEE; 2013. P. 3476–3483. https://doi.org/10.1109/CVPR.2013.446

10. Li Z, Hou Z, Chen C, Hao Z, An Y, Liang S, et al. Automatic Cardiothoracic Ratio Calculation with Deep Learn-ing. IEEE Access. 2019;7:37749–37756. https://doi.org/10.1109/ACCESS.2019.2900053

11. Hansson K, Häggström J, Kvart C, Lord P. Interobserver Variability of Vertebral Heart Size Measurements in Dogs with Normal and Enlarged Hearts. Veterinary Radiology and Ultrasound. 2005;46(2):122–130. https://doi.org/10.1111/j.1740–8261.2005.00024.x

12. Titov AD, Demichev VV, Akchurin SV, Shmarenkova YuS. Formation and Markup of a Specialized Dataset for Computer Vision Tasks in Veterinary Radiography. Vestnik of Russian New University (Vestnik of RosNOU). Series: Complex Systems: Models, Analysis and Management. 2025;4:129–140. (In Russ.). URL: https://vestnik-ros-nou.ru/item/2025/4/129

13. Vasilev ME, Shalimov AS, Savina OA. Overview of Yolo Versions: Single-Stage Convolutional Neural Network Model. Universum: Technical Sciences. 2025;6(135):36–45. (In Russ.). URL: https://7universum.com/ru/tech/ar-chive/item/20293

14. Hennessey E, DiFazio M, Hennessey R, Cassel N. Artificial Intelligence in Veterinary Diagnostic Imaging: A literature Review. Veterinary Radiology & Ultrasound. 2022;63(Suppl. 1):851–870. https://doi.org/10.1111/vru.13163

15. Levicar C, Nolte I, Granados-Soler JL, Freise F, Raue JF, Bach JP. Methods of Radiographic Measurements of Heart and Left Atrial Size in Dogs with and without Myxomatous Mitral Valve Disease: Intra- and Interobserver Agree-ment and Practicability of Different Methods. Animals. 2022;12(19):2531. https://doi.org/10.3390/ani12192531. Correc-tion in: Animals. 2025;15(8):1082. https://doi.org/10.3390/ani15081082

16. Cruz Rivera S, Liu X, Chan AW, Denniston AK, Calvert MJ; SPIRIT-AI and CONSORT-AI Working Group. Guidelines for Clinical Trial Protocols for Interventions Involving Artificial Intelligence: the SPIRIT-AI Extension. Lan-cet Digital Health. 2020;2(10):e549-e560. doi: https://doi.org/10.1016/S2589-7500(20)30219-3