Quantitative Analysis of Voice in Parkinson Disease Compared to Motor Performance: A Pilot Study

INTRODUCTION

The most prominent feature of voice disturbance in PD is hypokinetic dysarthria, characterized by monopitch and monoloudness [1]. Vocal changes in PD share similarities with other features of motor impairment found in PD, such as decreased dexterity and decay in repetitive actions [2–4]. Previous studies have reported that acoustic characteristics of speech and voice impairment which differ qualitatively among various neurological disorders, may offer a means for detecting early disease specific changes that might assist in monitoring disease progression [5–13]. Past studies have explored the relationship between speech changes and motor function in PD [12, 13]. In one study, improved bradykinesia and rigidity resulted in improved vowel articulation and pitch flexibility when participants were on-medication suggesting that the improved flexibility of physical movements, which occurs with dopaminergic medication, have positive effects on lip, tongue and laryngeal function by lessening postural tension and allowing for greater range of oral motor and laryngeal muscle movement [13]. The authors proposed that monitoring speech production may be a valuable marker of disease progression in PD [13]. Another study reported strong correlations between speech and axial and non-axial motor symptoms in nine individuals with PD when tested on-medication, possibly indicating that speech may be related to both dopaminergic and non-dopaminergic lesions in PD, and that speech production may serve as a marker of disease progression and medication treatment efficacy in PD, although the sample size was small [12]. The prevalence of voice abnormalities while on-medication has been documented in individuals with both lower and higher UPDRS scores [14]. Articulation and fluency problems were also noted in individuals with longer duration of PD in addition to voice problems, possibly due to initial neural mechanism changes in the brainstem in PD [14, 15]. Linguistic and cognitive changes, which result in fluency and prosody abnormalities, may occur as the disease progresses upward to higher order associations [14, 15].

When assessing acoustic parameters of voice, the features that have been most studied measure the stability of the vocal signal including, cycle-to-cycle variations of frequency of the vocal signal (“jitter”), cycle-to-cycle variations of amplitude of the vocal signal (“shimmer”), signal-to-noise ratio (SNR), and maximum phonation frequency range (MPFR), which is an indication of vocal flexibility [16, 17]. These measures are thought to reflect involuntary physiologic changes of frequency and intensity of vocal fold movement in contrast with voluntary changes in pitch and loudness [16, 18–23]. Coordinated vocal fold vibration with an intact phonatory system is reflected in low jitter and shimmer scores, and high signal-to-noise ratio scores [16, 20].

Individuals with PD have demonstrated a decrease in jitter after levodopa administration, indicating that this medication may improve vocal stability [24]. However, other studies have shown aberrant vocal acoustic findings regardless of medication treatment, but these reports did not quantify Parkinsonian features [25, 26]. Past studies have also reported a relationship between declining vocal function and disease severity however, only one trial of sustained phonation was examined and medication effects were either not controlled or subjects were assessed while on anti-Parkinson medication [27–29]. There has also been an attempt to predict stages of vocal decline over time based upon acoustic voice parameters and estimated growth curves for learning in individuals with PD [29]. Results estimated declines in fundamental frequency range and frequency variability over one to three years however subjects were tested on-medication and were not staged for severity of Parkinsonism [29]. Since these reports did not quantify Parkinsonian features, or control for medication effects, the relationship between the pattern of vocal change and disease severity in PDis unclear.

The aim of this study was to determine if selected acoustic parameters of voice; jitter, shimmer, signal-to-noise ratio and maximum phonation frequency range (MPFR) may be sensitive indicators of disease progression in PD and to evaluate correlations between acoustic parameters of voice and subtests of the UPDRS in individuals with milder and moresevere PD.

METHODS

Levodopa-responsive participants with typical diagnostic features of PD were divided on the basis of disease severity using the summed ADL section(Part 2) of the UPDRS, (which can range from 0 to 52 points among the 13 questions rated 0-4). A median split of scores defined the “mild” and “moderate” PD groups. Separating the groups in this manner provided a means for ranking on the basis of disability-related impairments typical of PD. The median split was at an ADL score of 12. Gender and age distribution is described in Table 1. All participants were non-smokers without a history of laryngeal or neurological surgery and none had voice problems unrelated to their neurological diagnosis. For all, hearing appeared functional as judged subjectively in conversation in the quiet room environment and a hearing screening revealed normal hearing in at least one ear. Twenty-three of the PD participants underwent videolaryngostroboscopy (VLS) to assess vocal fold vibration and to rule out any vocal fold pathology unrelated to their dysarthria.

DATA COLLECTION PROCEDURES

An AKG C-410 hypercardioid headband microphone with windscreen attached, (frequency range 20–20 kHz) was placed on the subjects and positioned 6.3 cm lateral to the subject’s mouth. The microphone had direct input into the computer as described in a similar study of acoustic analysis in neurological disease [10]. The CSpeech waveform analysis program was used for acoustic analysis [30]. A sound level meter was used to provide feedback of vocal intensity during testing [31]. The sound level meter was set at 60–70 dB (C gain SPL) depending upon the natural intensity level of the subject. Subjects were asked to look at the volt unit intensity meter on the sound level meter during phonation in an attempt to keep within a 5 dB range during phonation. The sound level meter was placed 38 cm from the subjects for easy viewing of the volt unit meter [10]. All PD participants had discontinued anti-Parkinson medication for at least 12 hours at the time of testing. Subjects were not tested if they felt ill or complained of fatigue.

After demonstration and practice, each subject phonated /i/ for 3 seconds at comfort pitch, high pitch (including falsetto) and low pitch (without glottal fry), for three trials per pitch [10]. In order to determine MPFR, subjects were asked to begin at their habitual pitch level and glide up to their highest pitch, including falsetto, for one trial on the sound /i/. Subjects were then asked to begin at their habitual pitch level and glide down to their lowest pitch for one trial on the sound /i/ [18].

ACOUSTIC ANALYSIS PROCEDURES

A steady portion of sustained vowels was used as the acoustic analysis sample for this study [18]. To reduce the effects of onset and termination of voice on frequency and intensity perturbation, the mid-section of each vowel was analyzed [17]. A tokenized analysis of the mid-one second segment of each sample of sustained phonation was used for acoustic measurement (a token refers to a non-overlapping 100 ms subinterval) [30]. In order to perform acoustic analysis, 10 tokens were chosen within the mid-one second of each trial of sustained phonation. For each token, pitch period and jitter were represented in milliseconds, fundamental frequency was reported in Hz, shimmer was represented as a percentage, and SNR was reported in dB. The results of each token included the mean score, the maximum and minimum scores, and the standard deviation value for each acoustic parameter [30]. The averaged result of the mean score was used for the statistical analysis portion of the study. When a signal was too short to divide into milliseconds, a non-tokenized analysis took place. During non-tokenized analysis, the selected segment was treated as one token [30, 32]. In order to calculate MPFR, the last 300 ms of each glide was selected for non-tokenized analysis [10]. The frequency ranges were then converted to semitones for reportingpurposes [10].

STATISTICAL METHODS

An analysis of variance (ANOVA) was used to examine overall group data. Of particular interest to us was the pairwise testing. This was done using Student’s t-tests for which significance was judged as p≤0.05. As gender might have an important role in voice dynamics, the overall analysis was followed by sub-analyses stratified by gender. In all analyses we followed the above by a test for a linear trend in the means. All correlations estimated were Pearson correlations. We evaluated if influential points were driving the results.

RESULTS

There was a near significant difference betweenthe three groups in semitone range (p = 0.057)(Table 1, Fig. 1). The test for linear trend was significant (p = 0.026). Pairwise comparisons indicated a significant difference in semitone range between the controls and moderate PD (p = 0.036). In an analysis stratified by gender an overall difference was noted for males for semitone range (p = 0.003) and the lowest frequency obtained during the glide down task (p = 0.001). For both of these variables the test for trends was significant (p = 0.002, p = 0.001 respectively). Pairwise testing identified significant differences in semitone range for males between controls and mild PD (p = 0.014) and controls and moderate PD (p = 0.005). Pairwise testing also identified significant differences for males between controls and mild PD and the controls and moderate PD in the lowest frequency obtained during the glide down task (p = 0.002, and p = 0.001, respectively) (Table 2).

The correlation study revealed that, as UPDRS speech and motor scores worsened for both groups, semitone range also diminished. When separated by gender, the males in both PD groups demonstrated significantly reduced semitone range with worsening speech and motor function scores on the UPDRS. Results further indicated that aberrant vocal fold frequency and amplitude function at high pitch phonation positively correlated with more severe symptoms of PD, particularly for females. Findings also indicated a positive correlation between low pitch shimmer and motor speech scores, and a negative correlation between signal-to-noise ratio and motor speech scores on the UPDRS in mild PD (Table 3).

The results of the VLS assessment revealed a few laryngeal abnormalities in the male PD group. One subject with mild PD had bilateral vocal fold edema, and two subjects with moderate PD had right vocal fold paresis (with normal glottic closure), and one subject had stiff vocal folds.

DISCUSSION

One of the goals of this study was to determine if specific acoustic parameters of voice might serve as sensitive indicators of PD and its severity in PD. Our data indicate that semitone range assessment may be one method of serving this role, particularly for males with PD. The significance of the linear trend indicated a progressively lower semitone range as disease worsened. The glide down task for males showed a steadily increasing value (as opposed to decreasing to reach lower notes) for worsening PD. Males in both PD groups had difficulty reaching as low a frequency as the control group likely contributing to the significantly reduced semitone range for the males with PD. This finding is consistent with the described hypokinetic dysarthria of Parkinsonism, of which monopitch is a classic hallmark of the PD voice [33]. Other studies have indicated higher model pitch and marked reduction of pitch variability in males with early PD compared to controls, however, medication status was not reported [27]. Several factors may contribute to the finding of monopitch in our study such as vocal fold paresis or stiff vocal folds, as observed in three of the male subjects with moderate PD. Previous studies have demonstrated that males with PD have a higher speaking fundamental frequency than controls; this factor may also play a role in the inability for males to produce lower frequencies [27, 33]. Other studies have reported reduced thyroarytenoid amplitudes during EMG testing contributing to hypophonia in PD as well as bowed vocal folds or vocal tremor [34–36].

There were no significant findings in semitone range in the female group. This may have been due to the small sample size of females in the groups or the large variability of responses of the females as indicated by the large standard deviation of responses for females in the moderate PD group. Results however, demonstrated a trend toward a more restricted semitone range as seen in the responses of the females in the moderate PD group. In this group, participants were no longer able to produce vocal glides as low or as high as the females in the control and mild PD groups possibly indicating the beginnings of laryngeal rigidity and impaired pitch variability [13].

The second goal of this study was to determine if vocal acoustic parameters correlated with motor findings in PD. The data shows that, with increased speech scores in the UPDRS motor examination, semitone range decreased, indicating reduced vocal flexibility with worsening speech skills. Semitone range was also noted to decrease as the combined UPDRS ADL and motor scores increased, reflecting reduced vocal flexibility with advancement of symptoms. Both females and males in the combined PD groups demonstrated that the higher the combined ADL and motor scores on the UPDRS, the greater the vocal acoustic perturbation of frequency and amplitude at high-pitch phonation. This was also seen in the speech section of ADL in the more severe PD group. Our findings may indicate impairment in the subtle variations of cricothyroid muscle movement needed to adjust to high pitch, further indicating an unstable phonatory system in PD. Findings of an unstable phonatory system at pitch extremes were also noted in the mild PD group as impaired cycle-to-cycle amplitude of vocal fold vibration at low pitch was significantly positively correlated with speech skills on the motor section of the UPDRS. Findings also revealed an increase in noise in the vocal signal as speech scores worsened, indicating impaired synchrony and coordination of vocal fold vibration in the mild PD group. It is possible that the phonatory demands required for pitch extremes produced instability in periodicity of vocal fold movement even during earlier stages of PD. Assessing pitch extremes may be an effective method for identifying aberrant vocal function in PD. The negative correlation we found between shimmer and bradykinesia for the combined male groups and the total PD group is unexpected and merits further investigation.

In the current study, vocal changes at pitch extremes in early PD were observed in males. This finding may support previous reports of the need for early voice intervention in PD however, further investigation is required with a larger number of age and sex matched participants. Previous investigations have shown improvement in the direction of normal articulation and vocal volume in individuals with mild-moderate PD immediately following Lee Silverman Voice Treatment (LSVT ^®) and improve vocal intensity and semitone range when tested in the clinic environment two years following (LSVT ^®) in patients with idiopathic PD [37, 38]. Studies have also demonstrated that, following LSVT ^®, an individual’s perception of their speech communication problem is improved [39]. Based on these reports, intensive voice therapy in the early stages of PD may help maintain vocal flexibility as the disease progresses thus preventing monotone and monopitch particularly since voice disturbances may be found in the early stages of PD [14]. However,individuals in the later, more severe stages of PD were not included in the previous studies and therefore, the long term effects of voice therapy are unknown.

Recent studies have shown that analysis of speech and voice via specific algorithms can predict disease severity of PD. One of these studies used telemonitoring to determine that acoustic voice measures of sustained phonation were able to differentiate individuals with PD from healthy controls [40, 41]. The results of the current study may indicate that testing frequency extremes, such as highest and lowest obtained pitch, and frequency range may also be sensitive indicators of disease progression in PD. Assessing vocal changes in PD may provide additional insight into general motor function during early stages of PD. There are also multiple factors to consider which may influence vocal function as part of the dysarthria in PD including deficits in internal cueing, planning and executing movements, attention to producing vocalizations and cognitive-linguistic skills [42, 43].

Based upon the results of the current study, we propose that early investigation of vocal acoustics, particularly semitone range, may offer another aspect of determining the presence of vocal abnormalities in early-moderate PD, in an efficient and non-invasive manner. The findings further support that monitoring acoustic changes in voice may provide an additional way to monitor disease progression, the effects of medication on the laryngeal system or the effects of voice therapy, in order to improve communication function in individuals with PD [13, 37–39].

In this study, the lack of significant differences between the two PD groups may be a function of the small sample size, the unequal distribution of males and females in the study, particularly the mild PD group which only had two females, or the possibility that the PD groups were too similar in their stages of disease progression. Significant differences in acoustic variations of voice may have been seen if a “severe” PD group had been added to the study. It is further possible that even though all participants passed a hearing screening in at least one ear, some subjects may have had an unrecognized hearing impairment which may have affected their ability to self-regulate vocal production [44]. However, viewing the volt unit meter may have assisted intensity control.

In summary, frequency range may be an acoustic variable sensitive to disease severity in PD, however, the results of the current study should be considered preliminary and warrants further investigation with a large sample size and age and sex-matched participants in early through late stages of PD. In this study, we arbitrarily defined disease severity by means of a cut-off between lower and higher summed ADL scores in our patient population. Future investigation might make use of other criteria for distinguishing between the various ways that PD severity can be quantified. Future studies may also consider detailed analyses of vocal fold vibration via VLS at various pitches and throughout disease progression of PD to further quantify vocal changes and motor function.

CONFLICT OF INTEREST FOR ALL AUTHORS

None.

FULL FINANCIAL DISCLOSURE FOR THE PAST YEAR

Alice K. Silbergleit: None

Peter A. LeWitt: Dr. LeWitt has served as a consultant or advisor for Acadia, Civitas, Concit, Depomed, Impax, Insightec, Intec, Ipsen, Knopp Biosciences, Kyowa Hakko, Lundbeck, Merck, Merz, NeuroDerm, Noven, Osmotica, Parkinson Study Group, Pfizer, ProStrakan, Teva, USWorldMeds, and XenoPort, and has received speaker honoraria from The International Parkinson’s Disease and Movement Disorders Society, Lundbeck, USWorldMeds, and the World Parkinson Congress. He is compensated for services as Editor-in-Chief of Clinical Neuropharmacology and also serves without compensation on the editorial boards of Journal of Neural Transmission, Translational Neurodegeneration, and Journal of Parkinson’s Disease. The Parkinson’s Disease and Movement Disorders Program that Dr. LeWitt directs has received clinical research grant support (for conducting clinical trial and other research) from Abbvie, Adamas, Addex, Allergan, Biotie, Great Lakes Neurotechnologies, The Michael J. Fox Foundation for Parkinson’s Research, Pharma 2B, Phytopharm, The Tremor Research Group, UCB, USWorldMeds, and XenoPort.

Edward L. Peterson: None.

Glendon M. Gardner: None.