Comparison of Two Auditory-Perceptual Evaluation Indexes, CAPE-V and GRBAS (Machine Learning), in Patients With...

This study aimed to compare the effectiveness of two auditory-perceptual voice assessment tools, GRBASZero (an AI-based application) and Consensus Auditory-Perceptual Evaluation of Voice (CAPE-V), in evaluating voice impairments in patients with Parkinson's disease (PD). The research sought to determine the sensitivity and reliability of these tools across different stages of PD severity, focusing on parameters such as overall severity, roughness, breathiness, strain, loudness, and pitch.

A cross-sectional study was conducted with 44 PD patients, categorized by disease severity using the Hoehn and Yahr scale. Voice samples were collected through sustained vowels /a/. Trained speech-language pathologists (SLP) performed CAPE-V assessments, while GRBASZero evaluations were conducted using an AI application. Statistical analyses included the Kruskal-Wallis test and correlation tests to compare the tools' performance.

This study showed that CAPE-V has higher sensitivity in identifying voice disorders at all stages of the disease (all P values < 0.05), while GRBASZero only detects Breathiness and Strain parameters at advanced stages (P = 0.049 and P = 0.034, respectively). Analyses were performed based on the vowel /a/ and the common parameters of the two instruments. The findings highlight the importance of choosing the appropriate instrument based on the stage of the disease.

CAPE-V is more comprehensive and sensitive for assessing voice impairments in PD patients across all disease stages, making it preferable for clinical and research use. GRBASZero, while useful for monitoring vocal strain and breathiness in advanced PD, lacks sensitivity in early stages. These findings highlight the importance of selecting appropriate tools for early diagnosis and monitoring of voice disorders in PD.