Using Machine Learning for the Automated Segmentation and Detection of Swallows Obtained by Digital Cervical Auscultation in Preterm Neonates.
Stephen So, Timothy Tadj, Belinda Schwerin and 3 others
PMID 40936063WHAT IT FOUND
A machine learning model identified 94% of swallows in preterm neonates from neck sounds.
It was more accurate for bottle feeds than breast feeds. This could help automate cervical auscultation, but it has not yet been tested against actual aspiration.
Key findings
01The model achieved 94% overall accuracy in identifying preterm neonate swallows from digital cervical auscultation recordings.
02Accuracy was higher for detecting swallows during bottle feeding compared to breastfeeding.
03The study lacked instrumental assessment to objectively verify the start and end points of swallows.
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
No instrumental assessment (like VFSS or FEES) was used to objectively verify the swallows, so the model's accuracy is judged against human listening. Inter- and intra-rater reliability for the manual segmentation was not formally reported. The model is less accurate for breastfed infants, likely due to the presence of non-nutritive sucking and dry swallows that were not manually segmented. The study validates swallow detection, not aspiration detection.
Declared interests
The study was funded by The University of Queensland.
The easy way to misread this
Do not assume this model detects aspiration. It only identifies the presence of a swallow sound. Using it to assess feeding safety without further validation against instrumental measures could lead to missed diagnoses of silent aspiration.
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 →