Construction of an Anchor and Training Sample Set for Auditory-Perceptual Voice Evaluation With the GRBAS-scale

Annelies Labaere; Marc De Bodt; Gwen Van Nuffelen

doi:10.1016/j.jvoice.2023.09.033

Construction of an Anchor and Training Sample Set for Auditory-Perceptual Voice Evaluation With the GRBAS-scale

J Voice. 2023 Nov 17:S0892-1997(23)00307-7. doi: 10.1016/j.jvoice.2023.09.033. Online ahead of print.

Authors

Annelies Labaere¹, Marc De Bodt², Gwen Van Nuffelen³

Affiliations

¹ Department of Translational Neurosciences, Faculty of Medicine and Health Sciences, University of Antwerp, Antwerp, Belgium; Department of Speech Pathology and audiology, Thomas More University of Applied Sciences, Antwerp, Belgium. Electronic address: annelies.labaere@thomasmore.be.
² Department of Translational Neurosciences, Faculty of Medicine and Health Sciences, University of Antwerp, Antwerp, Belgium; Department of Rehabilitation Sciences, Faculty of Medicine and health sciences, Ghent University, Ghent, Belgium.
³ Department of Translational Neurosciences, Faculty of Medicine and Health Sciences, University of Antwerp, Antwerp, Belgium; Department of Otorhinolaryngology, Head and Neck Surgery, Antwerp University Hospital, Antwerp, Belgium; Department of Rehabilitation Sciences, Faculty of Medicine and health sciences, Ghent University, Ghent, Belgium.

PMID: 37981533
DOI: 10.1016/j.jvoice.2023.09.033

Abstract

Objective: The aim of this study is to construct a representative and comprehensive set of reliable anchor and training voice samples for all parameters and grades of the GRBAS-scale, in order to develop a standardized online training tool for perceptual voice evaluation.

Methods: 140 voice samples varying from normal to severely dysphonic, consisting of the vowel /a/ and a Dutch standard text, were recorded. Samples were duplicated and added to a data folder in random order. A panel of five highly experienced judges rated GRBAS parameters for each voice sample. Intra- and interrater reliability were determined using intraclass correlation coefficient (ICC) and percentage agreement. Samples with agreement percentages of >0.90 for at least one parameter were qualified as anchor voices, whereas samples with a percentage agreement between 0.70 and 0.80 were considered valid training samples.

Results: Intrarater reliability was moderate to good for all raters and parameters. Interrater reliability was good for parameter G and moderate for R, B, A and S. 83 samples were found to be valid anchor samples with 90-100% expert agreement for at least one parameter, covering all of the GRBAS parameter/score combinations. An additional 37 samples fulfilled the criteria for adoption in the online tool as a training sample, showing 70-80% agreement for one or more parameters.

Conclusion: Based on well-defined and stringent selection criteria, this study identifies a substantial number of valid anchor and training voices that qualify for implementation in a standardized multi-level training tool for the auditory-perceptual evaluation of voice by means of the GRBAS scale.

Keywords: Auditory anchors; Auditory perceptual evaluation; GRBAS; Reliability; Voice quality.