Article | Jun 26, 2023
Aortic stenosis published in JASE. First study of its kind demonstrating the capacity of AI Echo to accurately quantify AoS severity with zero human input beyond image acquisition.
Key Findings
- This is the first study to apply a fully automated AI workflow to assess all degrees of aortic stenosis severity with zero human input beyond image acquisition, published in JASE alongside an editorial recognising it as a landmark for hands-off AS evaluation.
- Across a cohort of 256 patients (mean age 67.6 years) spanning normal aortic valves to severe AS, AI and human measurements of key Doppler and area parameters were closely matched, demonstrating that artificial neural networks can mimic expert echocardiographer performance.
- All 256 echocardiograms yielded at least one AI measurement (Vmax, mean pressure gradient, and/or AVA); 86% of studies yielded a complete AVA calculation, with the 14% exclusion rate driven by image quality rather than algorithm failure.
- The study validates the Us2.ai platform across the full AS severity spectrum, establishing a reference for automated AS grading that could reduce the workload of highly trained sonographers in busy valve clinics.
- As the first of its kind, the findings open the path to fully automated, consistent AS severity classification in clinical practice and large-scale research cohorts without the need for post-acquisition manual measurement.
AI Aortic Stenosis Assessment (publication)
Krishna, H., Desai, K., Slostad, B., Bhayani, S., Arnold, J. H., Ouwerkerk, W., Hummel, Y., Lam, C. S. P., Ezekowitz, J., Frost, M., Jiang, Z., Equilbec, C., Twing, A., Pellikka, P. A., Frazin, L., & Kansal, M. (2023). Fully Automated Artificial Intelligence Assessment of Aortic Stenosis by Echocardiography. Journal of the American Society of Echocardiography: Official Publication of the American Society of Echocardiography, 36(7), 769–777. https://doi.org/10.1016/j.echo.2023.03.008