Linear versus nonlinear modeling of dysphonia severity: A comparison between Acoustic Voice Quality Index and machine learning

Ahmed M. Yousef; Adrián Castillo-Allendes; Mark L. Berardi; Juliana Codino; Adam D. Rubin; Eric J. Hunter

doi:10.1016/j.jcomdis.2026.106648

Back

Linear versus nonlinear modeling of dysphonia severity: A comparison between Acoustic Voice Quality Index and machine learning

Journal article

Peer reviewed

Linear versus nonlinear modeling of dysphonia severity: A comparison between Acoustic Voice Quality Index and machine learning

Ahmed M. Yousef, Adrián Castillo-Allendes, Mark L. Berardi, Juliana Codino, Adam D. Rubin and Eric J. Hunter

Journal of communication disorders, Vol.121, 106648

05/01/2026

DOI: 10.1016/j.jcomdis.2026.106648

PMID: 42086020

View Online

Abstract

Purpose This study compares the Acoustic Voice Quality Index (AVQI-3) with machine learning (ML) models to evaluate their clinical utility for estimating voice quality. Methods Audio from 187 American English speakers (49 healthy, 138 with voice disorders) was rated for overall voice quality by six voice specialists. AVQI-3 and its six acoustic parameters were extracted, and these parameters were used to train 14 ML models (linear, curvilinear, nonlinear). Correlations and classification accuracy (normal–mild vs. moderate–severe) were compared against perceptual ratings as ground truth. Results AVQI-3 correlated strongly with perceptual ratings (Spearman =0.75), comparable to linear and curvilinear models, yet outperformed most nonlinear models. At a cutoff score of 2.52, AVQI-3 achieved the highest classification accuracy (0.92) with balanced sensitivity (0.90) and specificity (0.93). Among ML models, linear regression performed best (=0.77, accuracy=0.92, sensitivity=1.0, specificity=0.89), whereas nonlinear models showed reduced performance (average =0.74, accuracy=0.87, sensitivity=0.95, specificity=0.85). Conclusion AVQI-3 is a simple, accessible index that quantifies voice quality as effectively as complex ML models. This is supported by the best-performing ML models being linear, indicating that linear combinations of acoustic measures are effective, accurate, and clinically interpretable, whereas added nonlinear complexity offered little benefit.

Machine Learning

Acoustic Voice Quality Index

Speech acoustics

Voice assessment

Voice disorders

Voice quality

Details

Title: Subtitle: Linear versus nonlinear modeling of dysphonia severity: A comparison between Acoustic Voice Quality Index and machine learning
Creators: Ahmed M. Yousef - Center for Systems Biology
Adrián Castillo-Allendes - University of Iowa
Mark L. Berardi - University of Iowa
Juliana Codino - Lakeshore Foundation
Adam D. Rubin - Lakeshore Foundation
Eric J. Hunter - University of Iowa
Resource Type: Journal article
Publication Details: Journal of communication disorders, Vol.121, 106648
DOI: 10.1016/j.jcomdis.2026.106648
PMID: 42086020
ISSN: 0021-9924
eISSN: 1873-7994
Publisher: Elsevier
Grant note: R01DC012315 / National Institute on Deafness and Other Communication Disorders (100000055)
Language: English
Date published: 05/01/2026
Academic Unit: Communication Sciences and Disorders; Teaching and Learning; Otolaryngology
Record Identifier: 9985161448102771

Metrics

1 Record Views

See more details