Barsties V Latoszek Ben, Maryn Youri, Gerrits Ellen, De Bodt Marc
Faculty of Medicine and Health Sciences, University of Antwerp, Antwerp, Belgium; Institute of Health Studies, HAN University of Applied Sciences, Nijmegen, The Netherlands.
Faculty of Medicine and Health Sciences, University of Antwerp, Antwerp, Belgium; European Institute for ORL, Sint-Augustinus Hospital, Antwerp, Belgium; Faculty of Education, Health & Social Work, University College Ghent, Ghent, Belgium.
J Voice. 2017 Jul;31(4):511.e11-511.e27. doi: 10.1016/j.jvoice.2016.11.017. Epub 2017 Jan 10.
The evaluation of voice quality is a major component of voice assessment. The aim of the present study was to develop a new multivariate acoustic model for the evaluation of breathiness.
Concatenated voice samples of continuous speech and the sustained vowel [a:] from 970 subjects with dysphonia and 88 vocally healthy subjects were perceptually judged for breathiness severity. Acoustic analyses were conducted on the same concatenated voice samples after removal of the non-voiced segments of the continuous speech sample. The development of an acoustic model for breathiness was based on stepwise multiple linear regression analysis. Concurrent validity, diagnostic accuracy, and cross validation were statistically verified on the basis of the Spearman rank-order correlation coefficient (r), several estimates of the receiver operating characteristics plus the likelihood ratio, and iterated internal cross correlations.
Ratings of breathiness from four experts with moderate reliability were used. Stepwise multiple regression analysis yielded a nine-variable acoustic model for the multiparametric measurement of breathiness (Acoustic Breathiness Index [ABI]). A strong correlation was found between ABI and auditory-perceptual rating (r = 0.840, P = 0.000). The cross correlations confirmed a comparably high degree of association. Additionally, the receiver operating characteristics and likelihood ratio results showed the best diagnostic outcome at a threshold of ABI = 3.44 with a sensitivity of 82.4% and a specificity of 92.9%.
This study developed a new acoustic multivariate correlate for the evaluation of breathiness in voice. The ABI model showed valid and robust results and is therefore proposed as a new acoustic index for the evaluation of breathiness.
嗓音质量评估是嗓音评估的主要组成部分。本研究的目的是开发一种用于评估呼吸音的新的多变量声学模型。
对970名发声障碍患者和88名嗓音健康受试者的连续语音和持续元音[a:]的拼接语音样本进行呼吸音严重程度的感知判断。在去除连续语音样本的无声段后，对相同的拼接语音样本进行声学分析。基于逐步多元线性回归分析建立呼吸音声学模型。基于Spearman等级相关系数（r）、几种接收器操作特征估计值以及似然比和迭代内部交叉相关性，对同时效度、诊断准确性和交叉验证进行统计学验证。
使用了来自四位专家的具有中等可靠性的呼吸音评分。逐步多元回归分析产生了一个用于多参数测量呼吸音的九变量声学模型（声学呼吸音指数[ABI]）。发现ABI与听觉感知评分之间存在强相关性（r = 0.840，P = 0.000）。交叉相关性证实了相当高的关联度。此外，接收器操作特征和似然比结果显示，在ABI = 3.44的阈值下诊断结果最佳，灵敏度为82.4%，特异性为92.9%。
本研究开发了一种用于评估嗓音中呼吸音的新的声学多变量相关指标。ABI模型显示出有效且稳健的结果，因此被提议作为评估呼吸音的新声学指标。
