The Study at a Glance

The research focused on how consistently clinicians could identify lung sliding in both B-mode and M-mode ultrasound images. The goal was to assess interrater reliability and see how factors like experience level, training, and confidence played into diagnostic accuracy.

Despite lung sliding being a foundational element in chest POCUS education, results showed only moderate interrater reliability—with an intraclass correlation coefficient (ICC) of approximately 0.44.

Even more notably:

- B-mode outperformed M-mode in specificity, accuracy, and precision. - M-mode did not enhance diagnostic consistency, even among seasoned providers. - Training level, usage frequency, and self-reported confidence had no significant effect on diagnostic performance.

What Does This Mean?

These results point to an uncomfortable truth: We may be overestimating how reproducibly we interpret lung sliding in real-world clinical settings.

While M-mode is often viewed as a helpful fallback when B-mode images are ambiguous, the data challenges that assumption. In fact, reliance on M-mode may offer a false sense of diagnostic confidence without the consistency to back it up.

A Role for AI?

One promising implication: Artificial intelligence may be especially well-suited to assist with ultrasound pattern recognition—particularly in M-mode's distinct barcode-like outputs.

As machine learning models continue to improve in medical imaging interpretation, this could be a perfect use case for assistive diagnostics, helping to reduce variability and improve outcomes across experience levels.

Full paper available here: https://ojs.library.queensu.ca/index.php/pocus/article/view/17807/12355 Co-authors: Hans Clausdorff Fiedler, Delaney Smith, Derek Wu, Robert Arntfield