Ultrasonography-Based Patellar Tendon Area Measurement: Comparability of Automated vs. Manual Segmentation

Abstract

Introduction & Purpose: The patellar tendon shows region-specific structural adaptations to loading and aging, making accurate assessment of its anatomical cross sectional area (ACSA) clinically relevant (Ito et al., 2023). Ultrasound offers a practical alternative to MRI, but manual analysis is time-consuming and operator-dependent, motivating automated solutions. While several openly available tools for the automatic segmentation of the anatomical cross-sectional area of muscle exist, there is no open-source and peer-reviewed tool for the patellar tendon. In this study, we tested a fully automatic approach for the segmentation of the patellar tendon ACSA in panoramic and non-panoramic ultrasound images.

Methods: Images were acquired at 25%, 50%, and 75% of the length of the patellar tendon from 30 participants (age: 46.87±6.03 years; BMI: 25.45±4.14 kg/m²). To assess measurement consistency, we evaluated intra-rater and inter-session reliability using manual segmentation of the ACSA (Ritsche et al., 2022). Additionally, we trained three neural networks with a dataset of 497 images (156 panoramic images) to compare the performance of manual segmentation with automatic segmentation on a test set of unseen images.

Results: Intra-rater reliability was found to be good with ICC of 0.804 (95% CI: 0.628-0.902), SEM equal to 0.05 cm² (0.03-0.07), and MAE of 0.05 cm² (0.04-0.07); while inter-session reliability was excellent with ICC of 0.980 (0.970-0.987), SEM equal to 0.02 cm² (0.02-0.02), and MAE of 0.02 cm² (0.01-0.02). Regarding the comparability of the best model (i.e., UNet3+) after the removal of erroneous predictions with manual analysis, the ICC was equal to 0.848 (0.702-0.914), SEM was 0.05 (0.04-0.07), and MAE was 0.05 cm² (0.05-0.06) with a small standardized mean difference of 0.53 (0.33-0.75). When applying the model, analysis times per image ranged between 0.302 and 0.414 s.

Discussion: Both intra-rater and inter-session analyses demonstrated good to excellent reliability for assessing patellar tendon ACSA, confirming the repeatability of manual segmentation. Intra-rater reliability was slightly lower than in muscle ACSA studies, likely due to the tendon’s smaller size, where minor inaccuracies have a larger relative impact. Still, error margins were sufficient to detect known sex-related differences, though potentially insufficient for subtle age-related changes (Carroll et al., 2008). Inter-session reliability was excellent and consistent with existing literature, likely aided by standardized acquisition and reference images. The UNet3+ model slightly underestimated ACSA, and removing erroneous predictions only modestly improved agreement, emphasizing that visually subtle segmentation errors may escape detection yet remain physiologically meaningful. Model performance was particularly dependent on image quality and tendon boundary visibility. Overall, these findings support the feasibility of automated ACSA segmentation, with UNet3+ providing the most stable performance and substantial time savings, though expert oversight remains essential and further validation across devices, operators, and longitudinal designs is needed.

Conclusion: The proposed approach enables fast and less operator-dependent patellar tendon ACSA analysis. Although some differences were observed between manual and automatic analysis, this tool, when applied with caution, could provide valuable support in both clinical and research settings.

References

Carroll, C. C., Dickinson, J. M., Haus, J. M., Lee, G. A., Hollon, C. J., Aagaard, P., Magnusson, S. P., & Trappe, T. A. (2008). Influence of aging on the in vivo properties of human patellar tendon. Journal of Applied Physiology (Bethesda, Md.: 1985), 105(6), Article 6. https://doi.org/10.1152/japplphysiol.00059.2008

Ito, N., Scattone Silva, R., Sigurðsson, H. B., Cortes, D. H., & Silbernagel, K. G. (2023). Challenging the assumption of uniformity in patellar tendon structure: Regional patellar tendon morphology and mechanical properties in vivo. Journal of Orthopaedic Research: Official Publication of the Orthopaedic Research Society, 41(10), Article 10. https://doi.org/10.1002/jor.25563

Ritsche, P., Wirth, P., Cronin, N. J., Sarto, F., Narici, M. V., Faude, O., & Franchi, M. V. (2022). DeepACSA: Automatic Segmentation of Cross-Sectional Area in Ultrasound Images of Lower Limb Muscles Using Deep Learning. Medicine and Science in Sports and Exercise, 54(12), Article 12. https://doi.org/10.1249/MSS.0000000000003010