Parkinson's disease or atypical parkinsonism? The importance of acoustic voice analysis in differential diagnosis of speech disorders

Abstract Introduction Speech disorder is a common clinical manifestation in patients with Parkinson's disease and atypical parkinsonian syndromes and tends to occur before the onset of the axial parkinsonian symptoms. Due to parkinsonian features that overlap those of Parkinson's disease, the differentiation of voice and a speech disorder is a challenge for clinicians primarily in the early stage of the disease. Methods Speech samples were obtained from 116 subjects including 30 cases of Parkinson's disease, 30 cases of progressive supranuclear palsy, 30 cases of multiple system atrophy, and control group consisted of 26 subjects. Differential diagnosis of dysarthria subtypes was based on the quantitative, acoustic analysis of particular speech components. Additionally, Voice Handicap Index questionnaire was taken into account to differentiate the severity of voice impairment in the study groups. Results Our results showed significant differences in the distribution of acoustic parameters between Parkinson's disease and atypical parkinsonian syndromes. A mixed type of dysarthria with a combination of hypokinetic, spastic, and atactic features has been found in patients with atypical parkinsonism. In patients with the clinical diagnosis of the parkinsonian variant of multiple system atrophy, ataxic components of dysarthria were observed. Patients with PD presented pure hypokinetic dysarthria. Some parameters may be used as a marker for the diagnosis of the initial stage of PD. Voice impartment was significantly more frequent and severe in atypical parkinsonism than in Parkinson's disease. Conclusion Acoustic voice analysis is a very sensitive and noninvasive tool, provides objective information for the assessment of different speech components, has the specific potential to provide quantitative data essential for the improvement of the diagnostic process, and maybe a useful instrument in the differential diagnosis of parkinsonian syndromes.


| INTRODUC TI ON
Differentiation between Parkinson's disease (PD) and the atypical parkinsonian syndromes, such as progressive supranuclear palsy (PSP) and multiple system atrophy (MSA), is difficult, especially in initial stages when the clinical picture may be similar.
Parkinson's disease is a chronic and progressive neurodegenerative multisystem disorder, which causes damage to the dopaminergic neurons in the basal ganglia. Basal ganglia dysfunction implies the appearance of clinical symptoms in the form of bradykinesia, muscle rigidity, resting tremor, and postural instability (Berardelli et al., 2013;Darley, Aronson, & Brown, 1969a). Motor symptoms can often be preceded by the presence of nonmotor signs diagnosed as mood, behavioral, and cognitive disorder even a few years before the onset of the disease (Poewe, 2008). Most of the PD patients develop speech abnormalities defined as hypokinetic dysarthria described by mono-pitch and mono-loudness, variable rate, imprecise articulation, speech dysfluencies, inappropriate silence, reduced stress, and harsh voice quality (Darley et al., 1969a;Darley, Aronson, & Brown, 1969b;Logemann, Fisher, Boshes, & Blonsky, 1978). The previous study showed some form of vocal impairment in the early stages of the disease (Ho, Iansek, Marigliani, Bradshaw, & Gates, 1999;Logemann et al., 1978;Rusz, Cmejla, Ruzickova, & Ruzicka, 2011), and prosody turned out to be the most often affected speech subsystem in early untreated PD patients (Rusz et al., 2011).
The involvement of the basal ganglia, corticobulbar pathways, and the cerebellum causes various symptoms of speech disorders.
Routine assessment relies essentially on the qualitative description of the symptoms, and the differential diagnosis of speech impairment in parkinsonian syndromes especially during early stages of the disease may pose many difficulties. The hearing-perceptual assessment allows the examiner to distinguish the characteristic PSP features, most importantly articulatory phenomena and speech functions. Palilalia (involuntary repetition of own syllables, words, or even whole phrases), echolalia (automatic repetition of vocalizations made by another person) and to a lesser extent stuttering are often observed although they are not pathognomonic. Moreover, in PSP patients, verbal expression is generally limited and nonspontaneous.
As confirmed by database search, the subjective assessment of speech disorders in PD, PSP, and MSA is well documented (Darley et al., 1969a(Darley et al., , 1969bHartelius et al., 2006;Kluin et al., 1993Kluin et al., , 1996Logemann et al., 1978) whereas studies reporting data obtained using quantifiable measurements and describing differentiating PD from atypical parkinsonian syndromes based on speech assessment are scarce (Huh et al., 2015;Rusz et al., 2015Rusz et al., , 2019Skodda, Gronheitet, & Schlegel, 2012;Warnecke et al., 2019). The aim of the study was the quantitative assessment of speech dimensions and differential diagnosis of dysarthria based on the voice acoustic analysis.  for PSP, and the "Second consensus statement on the diagnosis of multiple system atrophy" for MSA (Berardelli et al., 2013;Gilman et al., 2008;Höglinger et al., 2017). To exclude other diseases, all subjects underwent a neuroimaging study (magnetic resonance imaging, MRI). The disease duration was estimated based on the self-

| ME THODS
Speech assessment (presence, type) was based on perceptual classifications of speech dimensions adapted from Darley et al. (1969aDarley et al. ( , 1969b. Frenchay Dysarthria Assessment (FDA) was used to assess speech abnormalities. The test is composed of eight categories in order: reflexes, respiration, lips, jaw, palate, larynx, tongue, and intelligibility subdivided into 28 specific vocal activities. Each of them consists of a 5-point scale, where '0' means no speech disorder, and the higher the score, the greater the speech disorder. Next, the Mayo Clinic dysarthria classification was used (Duffy, 2005). The level of voice and speech disability was also measured using the patient self-assessment questionnaire, the Voice Handicap Index (VHI) (Jacobson et al., 1997). A patient assesses voice impairment in three domains: functional, physical, and emotional. Each domain consists of 10 items (30 items in total), assessed on a scale of 0-4. The total VHI score reflects the severity of voice impairment as perceived by the patient and falls within the following ranges: a score of 0-30 points indicates mild dysphonia, of 31-60 points-moderate dysphonia, above 61 points-severe dysphonia (Jacobson et al., 1997;Pruszewicz, Obrębowski, Wiskirska-Woźnica, & Wojnowski, 2004).
The total VHI scores (VHI TOTAL) were taken into consideration to differentiate the severity of voice impairment in the study groups. In APS patients mixed dysarthria with hypokinetic, spastic, and ataxic components were diagnosed, and a range of severity was from moderate to severe. In most PD patients, hypokinetic dysarthria was observed. Severity has ranged from mild to moderate.
The acoustic voice analysis was performed in all subjects using the dedicated DiagnoScope Specialist software (DiagNova Technologies). Each task was done fully automatically by DiagnoScope Specialist software. Voice samples were recorded in a soundproof laboratory, with an average noise level of maximum 30 dB, using a large membrane multidirectional microphone, with bandwidth of 40 Hz-18 kHz, sensitivity 10 mV/pa, threshold sound pressure level 142 dB, and dynamic range 119 dB. During the recording, the microphone was on a support stand, at the level of the patient's mouth, at a distance of 20 cm (±5 cm). The acoustic signal was digitally processed using a 24-bit preamplifier M-Audio M-Track and saved on the hard drive of a computer running on a 64-bit operating system. The acoustic assessment has covered the voice performance module that allows evaluating voice performance and the phrase analysis module that allows evaluating words intonation. Participants have been recorded during a single session. All subjects were instructed to perform two vocal tasks, sustained phonation of the vowel/a/ as long as possible per one breath repeated three times. Only the best acoustic performance of the sample was considered. Second task phrase analysis (reading), participants were asked to read sentences at their usual rate and loudness. Phrase analysis included nine sentences that have differed intonationally. It has contained three sentences in the questioning, indicative, and imperative mood. The reading task was the same for all subjects, and the reading order was also the same.

| Statistical analysis
Statistical analysis was performed using the STATISTICA 13.1 package. Normality of distribution was assessed with the Shapiro-Wilk test. The normal-like distribution was analyzed using the one-way ANOVA. The distribution data deviating from normality were analyzed using the Kruskal-Wallis test. Post hoc analyses were conducted using the Tukey mean difference test for the one-way ANOVA and multiple comparisons of mean ranks for all samples tested by the nonparametric Kruskal-Wallis. Bonferroni's adjustment for multiple comparisons was used, and a significance level was set at p < .0125.
Differences in the distribution of the qualitative data were analyzed using the Chi-square test.

| Ethical compliance statement
Approval for this work was obtained from the Ethics Committee of the Medical University of Warsaw, KB/6/2016. Each participant provided written, informed consent.

| RE SULTS
Demographic and clinical data of the patients are shown in Table 1.
PD subjects were significantly younger than PSP (p = .00001) and MSA (p = .027) subjects and CG (p = .019), but their disease duration was significantly longer compared with PSP (p = .0000) and MSA subjects (p = .0000; Table 1). Global cognition was significantly decreased in PSP patients compared to PD (p = .0000) and MSA groups (p = .0013). In the perceptual assessment, speech disorders were found in all patients, but in PSP and MSA patients were significantly increased compared to PD (p = .0000). The motor scores were significantly worse in patients with PSP compared to PD (p = .0122), and this difference was not observed in patients with MSA.
The results of the assessed acoustic parameters in the groups are shown in Table 3 CG, p = .0000) were statistically significant.
The VHI TOTAL was used to assess of the severity of voice impairment in the groups (Figure 1) TA B L E 2 Selected voice acoustic parameters and corresponding speech characteristics in hypokinetic, spastic, and ataxic dysarthria (Huh et al., 2015;Rusz et al., 2011Rusz et al., , 2015Rusz et al., , 2019Skodda, Grönheit, Mancinelli, & Schlegel, 2013)   .0000 Voice self-assessment VHI Total .0000 .0000 .0011 Note: F₀ dev standard deviation of fundamental frequency converted to semitone scale, E base period energy averaged over the length of the entire sample, voice intensity, MPT duration of sustained vowel phonation, Perf Coef phonatory efficiency, Jitter micro perturbations of frequency, Shimmer microperturbations of amplitude, NHR amount of noise in the speech signal, F₀ModDep depth of fundamental frequency modulation, vocal tremor, Shimm dev alterations of loudness in the prolonged phonation period, S2H subharmonic to harmonic ratio, BreaksCoef continuous intervals below the phonation threshold within the intervals denoted as phonation, NoPhonCoef ratio of total length of phonation to the maximum phonation time.
Hypokinetic parameters were reduced in each patient group but did not differentiate between these groups. According to Darley et al., hypokinetic dysarthria reflects extrapyramidal dysfunction (Darley et al., 1969a). Moreover, the hypokinetic components of dysarthria have strongly correlated with neuronal loss and gliosis in the substantia nigra pars compacta (Kluin et al., 2001). It is suggested that hypokinetic speech disorders are related to general motor impairment observed in PD or APS patients (Kluin et al., 2001;Rusz et al., 2019). According to the previous study, there is a relationship between the severity of speech disorder and the degree of general motor disability in patients with parkinsonian syndrome. (Huh et al., 2015;Kluin et al., 2001;Midi et al., 2008;Rusz et al., 2019;Skodda et al., 2011). Based on these results, we suppose that it may have a direct impact on the respiratory, phonatory, and articula- subglottal pressure, resulting in the air "escaping" through the glottis and turbulently flowing through the vocal tract, hence, decreasing the sound volume and intensity (Pinto, Chan, Guimarães, Rothe-Neves, & Sadat, 2017;Sachin et al., 2008;Skodda et al., 2011). Voice abnormalities expose features of PD speech related to phonatory impairment and are also confirmed by some researchers (Fox & Ramig, 1997;Rusz et al., 2011). In the current study, authors show that disturbances of acoustic energy distribution in PD patients compared to controls can be attributable to bradykinesia and rigidity of intrinsic laryngeal muscles. Our theory was supported by the results of the previous study (Huh et al., 2015;Midi et al., 2008;Perez, Ramig, Smith, & Dromey, 1996;Warnecke et al., 2019). The results have shown that PD-related dysphonia may reflect the rigidity of laryngeal muscles. Patients with PD had more severe changes in voice quality compared to controls, which were linked with larynx tremor and incomplete glottic closure. This phenomenon is common in 60% of PD patients, and the mechanism was associated with hypokinesia and rigidity of laryngeal and/or respiratory muscles (Midi et al., 2008). In other study, authors concluded that vocal fold tremor is a typical clinical feature in most of the PD patients and is caused by vertical laryngeal tremor during vocal tasks (Warnecke et al., 2019).
Thus, we assume these parameters could be a useful marker for the diagnosis of PD in the early stage of the disease. Harsh voice quality was found in all patients but did not differentiate between groups of patients. It was markedly increased compared to patients and controls, except for Shimmer and NHR in PD, and NHR in MSA. Similar observations were found by Huh et al. (Huh et al., 2015). Speech is constrained, and forced, voice has a strained-strangled quality and is hoarse, and phonation unstable with breaks. Speech disturbances in our PSP patients were consistent with the results of the previous study that used objective measurements (Huh et al., 2015;Rusz et al., 2015Rusz et al., , 2019.
In MSA, speech disorders are described as mixed dysarthria MPT and uncontrollable changes in the voice pitch and volume allows differentiating between MSA and PD patients. The speech of MSA patients is characterized by hypophonia, dysprosody, and more nonsymmetrical mobility of the vocal folds than in PD patients, which in combination with the more pronounced changes in the voice pitch give an acoustic impression of a more severe vocal tremor. Based on a previous study (Warnecke et al., 2019), authors found that MSA patients exhibited laryngeal and pharyngeal disruptions on flexible endoscopic evaluation. All patients showed vocal fold abduction, resulting in a narrow glottic gap. Further, authors suggest that irregular arytenoid cartilage movements can be a biomarker for differentiation of MSA and PD (Warnecke et al., 2019). In another study, bilateral vocal fold motions impairment was found in 17 MSA patients, most of whom had moderate to severe bilateral vocal fold abductor restriction (Higo, Tayama, Watanabe, & Nitou, 2003).
Additionally, greater phonatory instability in the form of phonation breaks and no phonation was observed in MSA patients than in PD patients. Secondary to these changes, speech clarity (intelligibility) was decreased to such an extent that their comprehensibility to those around them was dramatically reduced. In the present study, such parameters as depth of fundamental frequency modulation or uncontrollable changes in the voice volume allowed distinguishing between MSA and PSP patients. The above observation is also confirmed by Rusz et al. (Rusz et al., 2015). They conducted the acoustic analysis of 12 patients with probable PSP, 13 MSA patients, and 15 PD patients. In objective assessment, speech disorders in APS patients seemed to be a combination of hypokinetic, spastic, and ataxic components. The speech of PSP patients (83% of subjects) was described as hypokinetic and spastic (hypokinetic in 51% and spastic in 43%) The speech of MSA patients was assessed as ataxic (in 56%) and spastic (in 45%), and speech disorders were mainly in the form of ataxic dysarthria (46%) or mixed dysarthria, with variable proportions of hypokinetic, ataxic, and spastic characteristics.

| CON CLUS IONS
Our study has limitations. First of all, groups are not gender-balanced. Therefore, a gender impact on the results cannot be excluded.
It is suggested that the discrepancy may be due to the anatomical structure of the larynx, and various weights of disease change on neural reflex for speech generation (Huh et al., 2015). On the other hand, an additional subanalysis of acoustic parameters in the control group was performed. The results clearly indicate that there are no differences in the assessed parameters between male and female.
The acoustic assessment was focused only on phonation and reading tasks. Further speech dimensions, for example, speech rate and rhythm, diadochokinesis, and articulation should be investigated.
Database search for studies on speech assessment in PD and APS patients found mainly studies focused on perceptual speech parameters. The acoustic analysis allows quantitative and objective assessment of voice parameters. It is a very sensitive, objective, and noninvasive tool, providing quantitative data essential for objective measurement of speech disorders; hence, it is a more accurate tool than perceptual assessment and may be useful in the differential diagnosis of parkinsonian syndromes. Moreover acoustic analysis can provide feedback of disease progression and treatment response. Thus, further research is needed using more accurate measurement scales to assess voice parameters in specific speech disorders to learn more about the neuropathology and mechanisms of their development in PD and APS.

CO N FLI C T O F I NTE R E S T
The authors declare no competing financial interests.

AUTH O R S CO NTR I B UTI O N
Renata Kowalska-Taczanowska, MSc involved in substantial contributions to conception and design, acquisition, analysis and interpretation of data, and drafting the manuscript. Andrzej Friedman,

MD, PhD, Chair and Professor of Neurology involved in substantial
contributions to conception and design, acquisition, analysis and interpretation of data, revising the manuscript critically for important intellectual content, and approval of the version to be published.

Dariusz Koziorowski, MD, PhD involved in substantial contributions
to conception and design, acquisition, analysis, and interpretation of data, revising the manuscript critically for important intellectual content, and approval of the version to be published.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data supporting the results in this study are available from the corresponding author upon reasonable request.