Automated assessment of hyoid movement during normal swallow using ultrasound.
Joan K-Y Ma, Alan A Wrench
PMID 35285113WHAT IT FOUND
A deep learning tracker estimated hyoid movement from ultrasound with good agreement to manual measures, outperforming a simpler shadow-based method.
It offers a potential route to faster, automated swallowing assessment, though it currently requires specific high-end hardware.
Key findings
01The deep neural network (DNN) tracker showed good agreement with manual tracking for maximum hyoid displacement, while the shadow tracker showed only moderate agreement.
02The DNN tracker also showed higher agreement with manual tracking for peak hyoid velocity than the shadow tracker.
03The DNN tracker was more robust and accurate than the shadow tracker, which required manual review and adjustment of regions of interest.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
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
The sample size was very small (15 participants), all of whom had normal swallowing. The study only assessed a 10 ml water bolus; results may not generalize to different consistencies or volumes. The DNN tracker was trained on data from a single probe and ultrasound system, so it may not work well with different equipment without retraining. The DNN tracker is not yet integrated into standard clinical software and requires specialized hardware to run in real-time. The training data for the DNN was included in the agreement calculations, which may inflate the apparent accuracy due to the small sample size.
Declared interests
One author is the director of Articulate Instruments Ltd, the company that manufactures the ultrasound system and software used in the study.
The easy way to misread this
Do not assume this automated tracking system is ready for immediate clinical use. The most accurate method requires specialized high-end hardware and was trained on a very small, healthy sample using a single type of ultrasound probe.