Negative binomial modeling of musculoskeletal ultrasound grayscale histograms: a three-device comparison and harmonization study.
Tomas I Gonzales, Katie L Boncella, Grace Begnell and 2 others
PMID 42466076WHAT IT FOUND
Muscle ultrasound brightness readings are not comparable between machines: the same muscle measured differently on each of the three scanners tested.
A statistical conversion brought brightness readings into agreement across devices, but texture converted less reliably, and the SonoSite Titan was hardest to convert.
Key findings
01The same muscle gave noticeably different brightness and texture readings on each of the three scanners, so a value from one machine cannot be read as if it came from another.
02The researchers built equations that converted mean muscle brightness from one machine into what another machine would have recorded, and those conversions agreed closely with the measured values.
03Converting texture (how spread out the grayscale values are) was less accurate than converting brightness, with prediction errors growing at the highest and lowest predicted values.
04Prediction bias did not differ significantly by muscle site, so the conversion did not work better for some muscles than others.
STILL TO COME
How it was doneWhat they foundWhat it means for PTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
Only 19 people were scanned, all of them older male U.S. Veterans who were mostly overweight or obese, so the conversion equations may not hold for women, younger or leaner people, or people with muscle disease. The authors say the participants' own tissue characteristics may have shaped the results. There was no separate group held back to test the equations on, and only one operator at one centre, so how well the conversions work in a new clinic, in different hands, or on a different day is untested. The equations are tied to these three machines. Another scanner would need its own version built from scratch, and they cannot be applied directly to a different device. All images were taken on each machine's factory default settings. Whether the conversion still holds when gain, time-gain compensation or dynamic range are changed is unknown, and the authors call this a critical area for future work. Ultrasound artifacts such as acoustic shadowing, reverberation, anisotropy and attenuation-related signal loss were not modelled or corrected, and images the operator judged unusable were left out of the analysis. The method works on processed B-mode images, so it harmonises the output of each machine's own image processing rather than the underlying signal.
Declared interests
Funded by the National Center for Advancing Translational Sciences at the National Institutes of Health, through the Clinical and Translational Science Awards Program (grant UL1TR000101). The authors declared that financial support was received for the work or its publication. No involvement by an ultrasound device manufacturer is stated.
The easy way to misread this
Do not read this as showing that muscle ultrasound values are now interchangeable between machines in practice. The conversion equations were built and cross-validated within the same 19 older male Veterans, were never tested on a separate group, and apply only to the three machines tested at their factory default settings, so the numbers will not carry over to a different scanner, a different patient, or a scan taken with adjusted settings.
Summarised by AI from the full paper, without a clinician reviewing it. Check it against the source before it changes what you do. Read it on PubMed →