JCPSLP vol 14 no 3 2012

Technology

Objective measurement of dysarthric speech following traumatic brain injury Clinical application of acoustic analysis Christine Taylor, Vanessa Aird, Emma Power, Emma Davies, Claire Madelaine, Audrey McCarry, and Kirrie J. Ballard

Speech pathologists typically use perceptual features and clusters of features to diagnose dysarthria type. Although ecologically valid, perceptual assessment remains largely subjective. This paper describes a sample of readily available acoustic measures and their perceptual correlates that can be applied in the clinical setting in order to objectively evaluate the degree of impairment and outcomes of intervention. The speech of three individuals with acquired dysarthria secondary to traumatic brain injury was perceptually rated for diagnosis. The samples were then analysed acoustically using measures that potentially quantify these perceptual features. Results indicated that most features were well quantified by an acoustic measure(s), while others were less clear. Some acoustic measures may be less sensitive to mild impairments while more extensive normative data are required for other measures. However, the acoustic measures used here provide a starting point to objectively describe dysarthric features, document treatment outcomes, and support accountability in service provision. D ysarthria is a disorder of speech motor control that affects one-third of individuals with traumatic brain injury (TBI) (Duffy, 2005). Dysarthria has a significant and sustained effect on quality of life. People with dysarthria have a reduced ability to communicate effectively in everyday activities, which can lead to social, vocational and life participation restrictions (WHO, 2001). The current gold standard for clinical diagnosis of dysarthria is subjective perceptual judgement of speech behaviours across a range of tasks. Perceptual measures are considered of highest value in terms of ecological validity (Duffy 2005). However, characterising dysarthria types can present challenges due to the inherent variability seen both within and across speakers. In addition, inter-rater agreement among non- expert clinicians on presence and severity of perceptual speech dimensions can be as low as 50–60% (Zyski &

Weisiger, 1987). Kent (1996) provided a comprehensive review of factors that undermine reliability of perceptual judgements within and across clinicians. For example, our accuracy of judgements is vulnerable to effects of drift over time, as one becomes more familiar with a client’s speech, as well as our level of expertise and familiarity with the possible range of severity. Several researchers have developed objective measure­ ment protocols to address problems with perceptual judgements but, generally, these have not made their way into routine clinical practice (Kent & Kim, 2003; Ludlow & Bassich, 1984; Murdoch, 2011). Barriers may include perceived or real difficulties with access to technical equipment, reduced expertise, entrenched clinical practices, and lack of time to collect and analyse objective measures. Also, it has been argued that some objective measures (e.g., vocal jitter or shimmer) may not correlate well with perceptual features (e.g., vocal roughness or harshness) (Bhuta, Patrick, & Garnett, 2004). One possible reason for a low relationship for some measures may be the use of nonspeech or quasi-speech tasks or simple word-level tasks to avoid the highly varied nature of connected speech. In the contemporary delivery of health care, where accountability is paramount, the use of objective measurements can strengthen our assessment methods and tracking of improvement. Understanding which measures have a strong relationship to perceptual features at all levels of speech production is critical to this endeavour. A comprehensive review of such measures is beyond the scope of this paper and several excellent overviews are already available (e.g., Kent, Weismer, Kent, Vorperian, & Duffy, 1999; Thompson-Ward & Theodoros, 1998). Instead, we will provide a brief overview of some acoustic measures developed for measuring vocal quality and prosody, features commonly affected in dysarthria. When evaluating vocal quality, one usually measures fundamental frequency ( f0 ) and intensity, and signal to noise ratios in a stable production task (e.g., sustained ah ) to capture features such as habitual pitch, hoarseness, and breathiness. Frequency measures quantify the rate, range, and variability of vocal fold vibration. Jitter and shimmer measure cycle-to-cycle change in frequency and amplitude, respectively, with elevated values thought to indicate pathology (Kent et al., 1999). High jitter values may correlate with perceived roughness (Colton, Casper, & Leonard, 2006; but see Bhuta et al., 2004). Harmonics-to- noise ratio (HNR) reflects abnormal vibratory characteristics of the folds and correlates with perceived hoarseness (e.g.,

Keywords ACOUSTIC ANALYSIS

This article has been peer- reviewed ASSESSMENT DYSARTHRIA TECHNOLOGY TRAUMATIC BRAIN INJURY

Christine Taylor (top), Vanessa Aird (centre) and Emma Power

129

JCPSLP Volume 14, Number 3 2012

www.speechpathologyaustralia.org.au