Classification of Hearing Status Based on Pupil Measures During Sentence Perception.

Area of Science:

• Auditory Neuroscience
• Human Auditory Perception
• Oculomotor Physiology

Background:

• Speech understanding in noise is challenging, particularly for individuals with hearing impairment (HI).
• HI listeners may alter listening effort allocation compared to normal-hearing (NH) peers.
• Pupil dilation responses may reflect these differences in effort.

Purpose of the Study:

• To assess the sensitivity of pupil measures to hearing-related changes in listening effort.
• To investigate how pupil measures differentiate between NH and HI listeners during speech perception in noise.
• To utilize a machine learning framework to rank pupil measures based on their sensitivity to hearing status, signal-to-noise ratio (SNR), and task response.

Main Methods:

• Collected pupil data from 32 NH and 32 HI listeners during an adaptive speech reception threshold test.
• Calculated various pupil measures including Peak Pupil Dilation (PPD), Mean Pupil Dilation (MPD), Principal Pupil Components (RPCs), and Baseline Pupil Size (BPS).
• Employed a machine learning classification framework to assess the ability to predict hearing status, SNR, and task response using pupil measures.

Main Results:

• A combination of pupil measures was required for accurate classification of hearing status, SNR, and task response.
• Established measures like PPD, RPC2, and BPS, along with novel measures RPC1 and RPC3, were key predictors.
• Classification performance varied depending on the factor being predicted and the specific pupil measures used.

Conclusions:

• Machine learning can effectively rank pupil measures by their sensitivity to factors influencing speech perception in noise.
• Pupil responses are sensitive to hearing status, SNR, and task demands, allowing for reasonable classification.
• Different pupil measures are affected differently by these factors, enhancing our understanding of their utility in auditory research.