Universidade de aveiro's voice evaluation protocol

Luis M. T. Jesus, Anna Barney, Ricardo Santos, Janine Caetano, Juliana Jorge, Pedro Sá CoutoInterspeech 2009

TL;DR

This paper introduces the Universidade de Aveiro's Voice Evaluation Protocol (UAVEP) for European Portuguese (EP), a standardized tool for speech-language therapists (SLTs) to assess vocal quality. A pilot study demonstrated good to excellent inter-rater reliability for most parameters, suggesting the protocol can improve diagnostic precision and communication among SLTs.

What Problem Does This Paper Address?

The assessment of vocal quality in speech-language therapy, particularly for European Portuguese (EP) speakers, has historically lacked standardization. While numerous protocols exist, many are developed for other languages, leading to inconsistencies in how SLTs evaluate and document vocal pathologies. This lack of a unified approach hinders objective diagnosis, treatment planning, and inter-professional communication. The authors aim to address this by developing and validating a comprehensive, standardized protocol specifically for EP.

Methods

The core of this paper is the development and pilot testing of the Universidade de Aveiro's Voice Evaluation Protocol (UAVEP). The protocol incorporates both established perceptual scales and novel parameters tailored for EP. Key components include:

  • Perceptual Scales: An EP version of the widely used CAPE-V (Consensus Auditory-Perceptual Evaluation of Voice) and the GRBAS (Grade, Roughness, Breathiness, Asthenia, Strain) scales. The EP CAPE-V was adapted with new sentences designed to elicit all EP vowels and specific phonetic contexts.
  • Comprehensive Parameter Assessment: Beyond standard perceptual scales, UAVEP includes assessments of:
  • Glottal attack
    • Respiratory support
    • Respiratory-phonotary-articulatory coordination
    • Digital laryngeal manipulation (assessing crepitation, rigidity, and pain)
    • Laryngeal tension
    • Vocal quality after laryngeal manipulation
    • Muscle tension (cervical, scapula, facial)
    • Posture (standing and seated)
    • Verbal articulation and related disturbances
    • Voice aerodynamics, respiratory endurance, and maximum phonation time (MPT)
    • S/Z ratio
    • Orofacial motricity
  • Reliability Study: A pilot study was conducted with 10 patients exhibiting vocal pathologies. Two experienced SLTs independently evaluated the patients using the UAVEP. Statistical analyses, including Spearman's correlation coefficient and Cohen's Kappa, were used to assess inter-rater reliability for ordinal, quantitative, and nominal variables.

Datasets

The study utilized a dataset of 10 patients with vocal pathologies. These patients were recruited from the Department of Otolaryngology of the Hospital de São João, Porto, Portugal. The paper does not introduce a new, publicly available dataset but rather uses this patient group to evaluate the reliability of the developed protocol.

Key Insights and Ideas

  • Standardization is Crucial: The lack of a standardized protocol for EP voice assessment is a significant barrier to precise diagnosis and effective communication among SLTs.
  • CAPE-V and GRBAS Adapt Well: The EP versions of CAPE-V and GRBAS demonstrated good to excellent inter-rater reliability, suggesting these established scales can be effectively adapted for the EP language.
  • Objective Parameters Enhance Reliability: Parameters that can be assessed more objectively, such as digital laryngeal manipulation (crepitation), scapula tension, and laryngeal tension, showed high correlation and agreement, likely due to being evaluated by touch and listening.
  • Familiarity Breeds Reliability: Parameters frequently assessed in clinical practice (e.g., overall severity, roughness, breathiness, grade) showed higher reliability, indicating that SLT familiarity with these concepts contributes to consistent evaluation.
  • Diagnosis is Highly Reliable: The diagnosis parameter showed maximum agreement, likely because SLT diagnosis is often supported by clinical findings.

Key Results

The pilot study revealed generally high inter-rater reliability for most UAVEP parameters:

  • CAPE-V: High and statistically significant correlation values were found for overall severity (p=0.964), roughness (p=0.834), breathiness (p=0.991), and loudness change (k=1.000). Resonance (k=0.608) and strain (p=0.659) showed intermediate correlation. Pitch change (k=0.500) had low but statistically significant correlation.
  • GRBAS: High and statistically significant values were observed for grade (p=0.815) and breathiness (p=0.846). Strain (p=0.662) showed intermediate and statistically significant agreement.
  • Other Parameters: Glottal attack (k=0.830), respiratory support (p=0.964), respiratory-phonotary-articulatory coordination (p=0.859), digital laryngeal manipulation (crepitation, k=1.000), vocal quality after manipulation (k=1.000), scapula tension (p=0.910), laryngeal tension (p=0.891), and diagnosis (p=1.000) all showed high correlation and statistical significance.
  • Lower Reliability: Variables like cervical tension (p=0.586) and facial tension (p=0.509) showed intermediate correlation with no statistical significance. Low correlation and no statistical significance were found for digital laryngeal manipulation (rigidity), and standing/sitting positions.

Limitations and Future Work

  • Limitations:
    • Small Sample Size: The study used a relatively small sample of 10 patients, which the authors acknowledge may influence the findings.
  • Non-Responses: A significant number of non-responses for some parameters limited the statistical analysis, particularly for quantitative variables in CAPE-V (pitch and loudness).
  • Future Work:
    • Expand Database: The authors are actively expanding their database to include approximately 100 patients to increase the robustness of the protocol's validation.
  • Further Validation: Future validation will involve creating a CD-ROM with voice samples for training and evaluation, including samples representing different severity levels (Mildly, Moderately, Severely Deviant).
  • Broader Application: The goal is to enable the protocol to be used with greater confidence in clinical practice and research.

Summary updated

BibTeX
@inproceedings{jesus2009universidade,
  author = {Luis M. T. Jesus and Anna Barney and Ricardo Santos and Janine Caetano and Juliana Jorge and Pedro Sá Couto},
  title = {{Universidade de aveiro's voice evaluation protocol}},
  booktitle = {Interspeech},
  year = {2009},
}
Figure