SLPOtherDysphagia2025

A Machine Learning Pipeline for Automated Bolus Segmentation and Area Measurement in Swallowing Videofluoroscopy Images of an Infant Pig Model.

Max Sarmet, Elska Kaczmarek, Alexane Fauveau and 9 others

PMID 40293507

WHAT IT FOUND

A free machine learning tool measured bolus area in infant pig swallowing videos with 97% less time than manual methods.

It achieved strong accuracy but required human review to correct errors caused by bubbles and surgical beads.

Key findings

01The machine learning pipeline reduced analysis time by 97% compared to manual measurement, taking 0.5 seconds per bolus versus 23 seconds.

02The model showed strong segmentation accuracy with a Dice similarity coefficient of 0.84 and an Intersection over Union of 0.76.

03Removing artifacts like bubbles and surgical beads through a human-supervised workflow improved the correlation between machine learning predictions and manual measurements.

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.

Already have one?

What it does not show

The study used an animal model (infant pigs), which allowed for higher frame rates (100 fps) than typical clinical settings (15–60 fps). This difference may affect how well the model performs on human videos. The automated method showed higher variability between raters (CV: 17.8%) compared to manual methods (CV: 5.3%). The model's accuracy was negatively impacted by artifacts such as bubbles and surgical beads, requiring a human-supervised step to remove them for best results. The study did not compare the model against a gold-standard volumetric measurement, but rather against manual 2D area measurements.

Declared interests

The authors declare that they have no conflict of interest.

The easy way to misread this

Do not assume this tool is ready for fully automated clinical use without human oversight. The study found that artifacts like bubbles significantly increased variability and required manual removal to achieve high accuracy, meaning the current pipeline is semi-automated rather than fully autonomous.

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 →