Body

Generalized perceptual Regression framework applied to Misophonia

Clonan et al. (2025) applied Generalized Perceptual Regression (GPR), a framework that maps a sound's statistical features onto a listener's behavioral ratings, to model misophonic sound aversion in 19 listeners (of 21) from the Free Open-Access Misophonia Stimuli (FOAMS) database who rated discomfort for 82 natural sounds and were assessed with the Duke Misophonia Questionnaire (DMQ) and the S-Five. Within GPR, each sound was represented by midbrain modulation power spectrum (mMPS) summary statistics derived from a cochlear and auditory midbrain model, reduced with principal component analysis, then fit with an individualized generalized linear model whose weights were projected back into the acoustic space to yield interpretable perceptual transfer functions for each participant. The authors concluded that GPR produces sound-computable, individualized models of misophonia that predict tolerance ratings, pinpoint the specific spectrotemporal features driving each listener's aversion, and yield gain and intercept parameters that correlate with established diagnostic scores, indicating potential for precision diagnosis (Clonan et al., 2025).