Combining Cardiovascular and Pupil Features Using k-Nearest Neighbor Classifiers to Assess Task Demand, Social Context, and Sentence Accuracy During Listening.
Bethany Plain, Hidde Pielage, Sophia E Kramer and 5 others
PMID 38549351WHAT IT FOUND
Physiological data can predict whether a hearing aid user feels observed, but only if the algorithm is trained on that specific person.
Group-wide models failed to predict listening effort or accuracy for new users, meaning current generic diagnostic tools cannot yet identify these states.
Key findings
01Classifiers trained on data from a single participant could predict social context (being observed) with high accuracy, but group-trained models failed to generalize to new participants.
02Predicting sentence accuracy (whether a repetition was correct) was the least successful task, with only two of 24 individual classifiers performing above chance.
03Cardiovascular features contributed more to the predictions than pupil features, and adding pupil data did not improve accuracy.
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 study used a laboratory task with artificial social pressure (observers rating the participant), which may not reflect real-world social anxiety. The group-level classifiers failed to generalize to new participants, meaning the initial high accuracy figures were misleading due to the way the data was split for testing. Sentence accuracy prediction was very poor, suggesting physiological signals do not reliably track moment-to-moment understanding in this context. The sample size was small (29 participants), and all were experienced users of a specific brand of hearing aids, limiting applicability to other devices or populations.
Declared interests
The authors declared no conflicts of interest. Funding came from the NIHR Manchester Biomedical Research Centre and the European Union's Horizon 2020 programme.
The easy way to misread this
Do not assume that physiological measures like pupil dilation can currently identify listening effort or accuracy in your patients. The study showed that while algorithms can predict these states when trained on a specific person's data, they fail completely when applied to a new patient, meaning no generic clinical tool is available yet.