Real-Time Echo Intelligence: Can AI Match or Beat Human Eyes?

/

Point‑of‑care echo is fast, powerful—and famously operator‑dependent. Most of us rely on quick visual impressions because precise measurements take time we often don’t have. This new study puts three automated, real‑time AI tools to the test, comparing their ejection fraction, VTI, and IVC assessments against an expert’s read. Join Dr. Sookdeo as she breaks down how well the algorithms performed and what this could mean for the future of bedside echo.

Gohar E, Herling A, Mazuz M, et al. Artificial Intelligence (AI) versus POCUS Expert: A Validation Study of Three Automatic AI-Based, Real-Time, Hemodynamic Echocardiographic Assessment Tools. J Clin Med. 2023;12(4):1352. Published 2023 Feb 8. doi:10.3390/jcm12041352.

Key Findings of the study:

  • The general idea: while manual views still must be obtained by a provider, some of the detailed measurements may be more accurate (and take less time) when obtained through artificial intelligence generated measurements.

  • Methodology: POCUS expert and AI tools made independent, blinded measurements on the same cardiac images

    • Three measurements were compared between POCUS expert and AI tools in the GE Venue device:

      • auto- left ventricular ejection fraction (LVEF)

      • auto- velocity time integral (VTI) (part of the data needed to calculate cardiac stroke volume and

      • auto- inferior vena cava (IVC) measurements

  • Results: AI showed good agreement with POCUS experts for high-quality images (Cohen's kappa: 0.498-0.655)

  • Critical limitation: Tools only worked well with good image quality - proper acquisition technique remains essential

  • Conclusion: "Garbage in, garbage out" - AI doesn't compensate for poor technique, but can perhaps be helpful in more accurate measurements with good views obtained

 Clinical Implications and other discussion points:

  • First-generation automated measurement tools require high image quality to function properly

  • Newer AI systems can extract diagnostic information from suboptimal views

  • AI-guided novices achieve 96% success in obtaining diagnostic images in under 5 minutes

  • Major potential for emergency departments, critical care, and resource-limited settings

 Limitations & Future Directions:

  • Study was small, single-center, and tested on expert-acquired images only

  • Workflow integration and medico-legal responsibility questions remain unresolved

  • Future: comprehensive AI systems that guide acquisition, interpret images, and predict outcomes

Take-Home Message:

  • Current automated tools are powerful assistants but don't replace proper technique and clinical judgment

  • AI should augment, not replace, clinical decision-making

  • Embrace tools thoughtfully while understanding their limitations

AUTHORSHIP

Written by: Maddie Sookdeo MD, PGY-3 University of Cincinnati Department of Emergency Medicine

Editing, Posting, and Audio Editing by Anita Goel, MD; Associate Professor, APD of UC EM Residency Program, and Co-editor of Tamingthesru.com

Cite as: Sookdeo, M., Goel, A. Real-Time Echo Intelligence: Can AI Match or Beat Human Eyes? TamingtheSRU.com. www.tamingthesru.com/blog/journal-club/ai-tools-for-bedside-echo. 5/4/26.