Multiparameter Voice Assessment for Voice Disorder

Patients: A Correlation Analysis Between Objective

and Subjective Parameters

Yu Ziwei, Pang Zheng, and Dong Pin, Shanghai, China

Summary: Objective. The aim of this study was to establish a multiparameter voice assessment profile using objec-

AccepFrom

UniversiAddre

gology, HHospital,Journa0892-1� 201http://d

tive multiparameter test and subjective voice quality assessment.Methods. We assessed 50 patients with voice disorders before and after operation. The assessment incorporates (1)subjective voice quality assessment, (2) patients’ self-assessment, and (3) objective acoustic analysis. The subjectivevoice quality assessment uses GRABS system to evaluates the grade of hoarseness (G), proposed by the JapaneseSociety for Logopedics and Phoniatrics. Patients’ self-assessment is modified based on the Chinese version of voicehandicap index (VHI) scale, composed of functional (F), physiological (P), emotional (E) part, and a total score (T).The acoustical analysis evaluate the patients’ voice sample by voice analysis software ‘‘Dr. Speech’’. Three parameters,jitter (J), shimmer(S), and normalized noise energy (NNE), were taken in analysis.Results. We observed high correlations among subentries F, P, and the total score TvH of the VHI scale in patients’subjective assessment. Parameter E does not correlate well with other assessed parameters. The Chinese versionof VHI, which incorporate multifactors including age, education, and especially the cultural difference may accountfor the inconsistent correction in parameter E. In the objective acoustic analysis, high correlation among the threeparameters J, S, and NNE is observed.Conclusion. Systemic assessment combining a subjective voice quality assessment, an objective acoustic analysis,and a self-assessment is helpful in clinical practice in the diagnosis and treatment for voice disorders. The E componentin VHI scale assessment may not be a reliable parameter to evaluate treatment outcome.Key Words: Voice disorder–Subjective voice quality assessment–Self-assessment–Acoustic analysis.

INTRODUCTION

Voice disorder is commonly seen in Otolaryngology patients.Disrupted voice function negatively impact patients’ psychol-ogy and social life, thereby affecting patients’ overall qualityof life. In 1997, Jacobson1 proposed the voice handicap index(VHI) scale, a self-assessment approach, to evaluate theimpact of voice disorders on patients’ physiological, social,and psychological functions. VHI score is widely used toevaluate patients’ quality of life under the condition of voicedisorders. It is a self-evaluation of the degree of voicehandicap and calculated as the sum of all questions (T) forthe following three domains: functional (F), physiological(P), and emotional (E) components to evaluate the impact ofvoice disorders on patients’ physiological function, socialself-adaptability, and emotional change. Self-assessment thusplays a key role in evaluating the degree of voice disordersand treatment outcomes. However, neither objective acousticalassessment nor morphologic assessment of laryngoscopecan reflect patients’ feeling of voice disorders on their lives,work and social activities, or the impact of voice disorderson mental health.

Lam et al2,3 reported formal testing of reliability and validityof the Chinese Hong Kong version of the VHI scale. Recently,

ted for publication March 20, 2014.the Department of Otolaryngology, Head and Neck Surgery, Shanghai Jiaotongty Affiliated First People’s Hospital, Shanghai, 200080, China.ss correspondence and reprint requests to Yu Ziwei, Department of Otolaryn-ead and Neck Surgery, Shanghai Jiaotong University Affiliated First People’s100 Haining Road, Shanghai 200080, China. E-mail: david545yu@yahoo.coml of Voice, Vol. 28, No. 6, pp. 770-774997/$36.004 The Voice Foundationx.doi.org/10.1016/j.jvoice.2014.03.014

Xu et al4 reported their work on the Chinese edition of VHIscale, showing good reliability and validity in assessment. InChina, the severity of voice disorder, evaluated in determiningtherapeutic effect and prognosis, has only recently been adop-ted by medical doctors. The assessment is based on the doctors’subjective opinion by hearing patients’ voice. Such judgment isindividual doctor-dependent, may lack a good reproducibility.With the development of computer-based assessment and med-ical research of voice, the objective voice detection methods arenow available. These objective methods are designed primarilyon voice detection of acoustics, aerodynamics, and physiolog-ical parameters.Auditory Perceptual Evaluation of Voice primarily used in

clinical settings is the most direct judgment used by clinicalphysicians and other professionals in evaluating patients’ voicequality. The Auditory Perceptual Evaluation of Voice aremainly based on two criteria: (1) the GRBAS scale (grade,roughness, breathiness, asthenia, and strain scale), proposedby the Japanese Society for Logopedics and Phoniatrics and(2) the CAPE-V scale, proposed by the American Speech–Lan-guage–Hearing Association. The GRBAS scale is the mostwidely accepted scale for voice evaluation in clinics world-wide.5–7 In China, subjective evaluations are the prevailingmethod and used only in large hospitals. The outcomeslargely depend on doctors’ clinical experience and assessmentskills.The integrated use of various test parameters to assess (quan-

tify) the quality of voice is still a debated yet important topic inclinical practice. Thus, the purpose of this studywas to establisha multiparameter voice assessment method to improve voicequality evaluation.

TABLE 1.

Comparison of VHI Parameters and Acoustic Analysis

Preoperative and 1Week After Surgery (Mean ± Standard

Deviation)

Variable Preoperative Postoperative P Value

VHI parameters

P 21.43 ± 3.24 4.02 ± 1.53 0.008

F 11.29 ± 1.76 3.77 ± 1.63 0.007

E 2.85 ± 1.44 1.54 ± 0.58 0.007

T 36.95 ± 3.05 8.02 ± 1.94 0.008

Acoustic

analysis

J 0.82 ± 0.31 0.19 ± 0.22 0.005

S 4.58 ± 0.14 1.42 ± 0.50 0.005

NNE �4.88 ± 2.13 �12.95 ± 2.01 0.006

All subentry P, F, E, and T scores of VHI were significantly reduced 1 week

after surgery (P < 0.01), and J, S, and NNE of acoustic analysis were

significantly reduced 1 week after surgery (P < 0.01).

Yu Ziwei, et al Objective and Subjective Analysis for Voice Disorder Patients 771

METHODS

Patients and therapeutic methods

Fifty voice disorder patients were included in the study fromthe period of June 2010 to October 2011. All patients wereadmitted to the Department of Otolaryngology, Head andNeck Surgery, Shanghai Jiao Tong University Affiliated FirstPeople’s Hospital. The average age of patients is 47.08 ± 3.94(standard deviation) years (range 25–70 years). Patients werediagnosed of vocal fold polyp (n ¼ 32), vocal fold nodule(n ¼ 5), vocal fold leukoplakia (n ¼ 5), vocal fold cyst(n¼ 5), and vocal fold Reinke edema (n¼ 3). For each patient,the removal of vocal fold lesion was performed under self-retaining laryngoscope, and inhalation treatment of PulmicortRespules (AstraZeneca Pty Ltd.) was applied for 1 week aftersurgery.

Instruments and voice assessment methods

Subjective voice quality assessment. All sound mea-surements were recorded in the acoustic room with a high-fidelity audio equipment (digital audio tapes). The distancebetween mouth and microphone was 10 cm. Each patientwas asked to read an assigned sentence in his/her naturaltone and loudness. The recorded materials from the digitalaudio tapes were then transferred to a computer. Three doctorswere assigned to independently evaluate the voice quality. Ac-cording to the grade of hoarseness (G) in GRABS system pro-posed by the Japanese Society for Logopedics and Phoniatrics,the voice quality was rated into one of four levels: 0 for normalvoice, l for mild hoarseness, 2 for moderate hoarseness, and 3for severe hoarseness. To minimize the assessment differenceamong the three doctors, thereby increasing the credibility ofassessment results, each doctor was trained to recognizetypical sound samples, and the subjects’ samples were alsorandomly arranged and presented to a given doctor three times.Each sound sample was thus assessed three times per doctorand then averaged.

Self-assessment. On the VHI scale, the impact of voice dis-order on patients’ quality of life is based on the assessment offunctional (F), physiological (P), and emotional (E) compo-nents. The total score of the three components is T and to sepa-rate from the impact of E, sum of F and P is used as TvH

(TvH ¼ F + P). Each component is evaluated based on 10 ques-tions. Patients were asked to rate each question based on the fre-quency of occurrence: 0 for never, 1 for seldom, 2 for sometimes,3 for regularly, and 4 for always. The score of each componentrange from 0 to 40, and the total score (T) is from 0 to 120.A higher score on a particular component indicates a greaterimpact of the voice disorder on this aspect for the patient; ahigher total score means a worse self-recognition of the patienton voice disorder.

Acoustic analysis. Acoustic test was conducted in an acous-tic room. A microphone was placed 10 cm away from patients’mouth. Subjects were then asked to pronounce the vowel /a/ for3 seconds. Sound samples were recorded and then transferredinto the computer for analysis using the voice analysis software.Acoustic analysis was conducted under the voice analysis

computer system (Tiger Electronics Co., Ltd) using the soft-ware Dr. Speech windows, version 4.0.8 Three parameterswere selected for analysis: jitter (J, cycle-to-cycle variation infrequency), shimmer (S, cycle-to-cycle variation in intensity),and normalized noise energy (NNE, relative level of vocal noiseto that of harmonics).

Statistical analysis

Matched t test was performed to analyze the differencesbetween preoperation and postsurgery. Spearman correla-tion was carried out to analyze the correlations among theparameters. All data analyses were performed with SPSSversion 13.0 (SPSS, Chicago, IL).

RESULTS

VHI scale assessment and acoustic analysis

The results of VHI scale assessment and acoustic analysisshow that both self-assessment and acoustic index significantlydecreased 1 week after surgery (P < 0.01) compared with pre-operative (Table 1).

Correlation analysis of subentries on VHI scale

Table 2 shows the correlation analysis results of self-assessment. Both before and after surgery, there were signif-icantly strong positive correlations between F, P, and the sumTvH (TvH ¼ P + F). Preoperatively, the Spearman correlation

TABLE 2.

Spearman Correlations Among Subentries of VHI Scale

VHI Parameter

Preoperative Spearman Coefficient Postoperative Spearman Coefficient

P F E T P F E T

F 0.843* 1.000 0.821* 1.000

E 0.282 0.340 1.000 0.195 0.296 1.000

T 0.552 0.482 0.283 1.000 0.610 0.502 0.277 1.000

TvH 0.931* 0.894* 0.262 0.356 0.897* 0.855* 0.301 0.364

* P < 0.01 for the test.

Journal of Voice, Vol. 28, No. 6, 2014772

coefficient between P and F was 0.843, between P and TvH

was 0.931, and between F and TvH was 0.894. Likewise,postoperatively, the Spearman correlation coefficient be-tween P and F was 0.821, between P and TvH was 0.897,and between F and TvH was 0.855. However, E and T werenot found in correlation with other index. This suggeststhat except for the emotional components, in both preopera-tive and 1 week after surgery, there were positive correla-tions among three subentries, that is, functionalcomponent, physiological component, and the total of thesetwo parts.

Correlation analysis of parameters in acoustic

analysis

All three parameters in the acoustic analysis are strongly corre-lated (Table 3). The pre- and post-operative Spearman correla-tion coefficient between J and S was 0.751 and 0.812,respectively; correlation coefficient between J and NNE was0.708 and 0.743, respectively; and correlation coefficient be-tween S and NNE was 0.653 and 0.703, respectively. This indi-cates that, preoperative and 1 week after surgery, there werepositive correlations among these three parameters, that is, J,S, and NNE.

Correlation analysis of subentries of VHI scale and

parameters in acoustic analysis

Results in Table 4 show that except for the emotional part, inboth preoperative and 1 week after surgery, there were positivecorrelations among all other components: functional, physio-logical, total of these two parts and J, S, and NNE, but notbetween E and T.

TABLE 3.

Spearman Correlations Among Acoustic Parameters

Acoustic

Parameters

Preoperative

Spearman

Coefficient

Postoperative

Spearman

Coefficient

J S J S

S 0.751* 1.000 0.812* 1.000

NNE 0.708* 0.653* 0.743* 0.703*

* P < 0.01 for the test.

Correlation analysis of subjective voice quality

assessment, subentries of VHI scale, and parameters

in acoustic analysis

There were strong correlations between the grade of hoarseness(G) and F, P, and TvH of VHI scale as well as J, S, and NNE inacoustic analysis (Table 5). This indicates that with the excep-tion of the emotional part, both preoperative and 1 week aftersurgery, there were positive correlations between the grade ofhoarseness and subentries of VHI scale and parameters inacoustic analysis.

DISCUSSION

Natural voice consists of various physical features. It is com-mon to use auditory perceptual evaluation as a referencemeasure for evaluating objective assessments.9 The GRBASsystem, developed by the Japanese Society for Logopedicsand Phoniatrics, is the prevailing auditory perceptualassessment. In this study, we included auditory perceptualevaluation in our voice assessment. The grade of hoarseness(G)10 provided the most reliable and stable results. Ourresults suggest that G can well reflect the severity of voicedisorders and serves as a surrogate to evaluate the severityof voice disorder, providing a good guidance for subsequenttreatment.Acoustic analysis as a method for objective assessment of

voice quality has been available in clinics. Parameter J in acous-tic analysis reflects the changes in sonic basic frequency be-tween the time point and the same time point of the adjacentcycle; S reflects changes in wave amplitudes between the pointand the same time point of adjacent cycle; and NNE reflects thenormalized noise energy due to the incomplete glottal closure.Specifically, J reflects the rough sound level; S reflects thehoarse sound level; and NNE reflects breath sound level, afterthe hoarse sound level. The selection of parameters in objectivevoice assessment should meet three fundamental principles: (1)must have pathophysiological significance and have someapplication ‘‘history’’; (2) changes in assessment results shouldbe consistent with that in the severity of voice disorders; and (3)normal and unusual voice can be effectively defined by valuesof the parameters. In this study, J, S, and NNE were noticeablyreduced 1 week after surgery compared with preoperative mea-surements. However, the clinical features, that is, rough voice,hoarse voice, and breath sounds, were reduced. These changesindicate that the voice quality in voice disorder patients was

TABLE 4.

Spearman Correlations Between VHI Scales and Acoustic Parameters

VHI Scale

Preoperative Spearman Coefficient Postoperative Spearman Coefficient

J S NNE J S NNE

P 0.841* 0.873* 0.885* 0.834* 0.742* 0.728*

F 0.835* 0.793* 0.680* 0.902* 0.834* 0.921*

E 0.210 0.243 0.265 0.181 0.203 0.295

T 0.329 0.531 0.412 0.527 0.582 0.423

TvH 0.593* 0.775* 0.654* 0.766* 0.732* 0.848*

* P < 0.01 for the test.

Yu Ziwei, et al Objective and Subjective Analysis for Voice Disorder Patients 773

noticeably improved after surgery. Therefore, acoustic analysesin this study provide objective quantification to evaluate surgi-cal outcome.

Voice disorders lead to psychological problem and social dif-ficulty, thus affecting patients’ quality of life. Subjective assess-ment using the VHI provides information in this aspect andtherefore serves as a complement to clinical and acoustic detec-tion. VHI as the voice self-assessment method possesses adegree of complexity that far exceeds those of objectivemeasures. Because of the diversity in personal factors, suchas social status, profession, social status, family situations, per-sonality, and so on, the impact of voice on patients’ life varies.11

In this study, the VHI scale was used as the tool for subjectiveassessment. Our original design planned to analyze the totalof E, F, and P as T. The experimental results showed poor cor-relation between E and the other parameters. There were strongcorrelations among the remaining parameters only when E wasexcluded from the analysis, and the total of F and P was re-corded as TvH. This suggests that the VHI scale, which basedon the patient’s subjective experience on impact of their voicedisorder, only effectively reflect the patients’ perception onthe functional and physiological aspects of their life, work,and social activities. The assessment is likely biased becauseof the traditional culture of China, where people in generalare reluctant to express their negative emotions selecteda lower score. Therefore, we found no significant changes

TABLE 5.

Spearman Correlations Between G and Both Subentries

of VHI Scale and Acoustic Parameters

Parameters

Preoperative_G

Spearman

Coefficient

Postoperative_G

Spearman

Coefficient

P 0.944* 0.954*

F 0.973* 0.920*

E 0.221 0.196

TvH 0.987* 0.965*

J 0.918* 0.939*

S 0.974* 0.968*

NNE 0.929* 0.972*

* P < 0.01 for the test.

in emotional scores before and after surgery. Answers to theemotional component of the VHI scale are also affected bythe patients’ age, occupation, education background, patientcompliance, and so on.12,13 The Chinese version of the VHIscale should therefore be revised to accommodate the easternculture background. We recommend that the emotional partshould be modified to be in line with the characteristic ofAsian culture.

Voice is a complex multidimensional phenomenon; patho-logic changes of vocal fold usually reduce the voice qualityin multiple aspects. Single parameter comes short to assessthe vocal function in both normal and abnormal conditions.Each available assessment method only evaluates one particularaspect of the laryngeal function. Voice assessment should there-fore include multiple parameters.14,15

CONCLUSIONS

In this study, we find that the assessment method is composedof subjective assessment, objective assessment, and a self-assessment helps to improve clinical diagnosis and treatmentof voice disorders. The use of a multiparameter voice assess-ment method is particularly helpful in the assessment in (1)the severity of voice disorders preoperatively, (2) evaluationof the treatment outcome, (3) comparing the outcomes ofdifferent treatment methods, (4) dynamic observation of thedevelopment and prognosis of disease, and (5) in patientfollow-up observations. Integrated use of subjective and objec-tive assessments can systematically evaluate and comparepatients’ voice quality pre- and post-operatively. The task inclinic is not only to remove lesions and clinically cure patientsbut also to care for patients’ physiological recovery and theirsocial adaptabilities as well. The multiparameter methods willhelp doctors design treatment targets that are consistent withthe patients’ subjective feelings and life requirements. Theassessment method introduced in this work is expected toplay an important role for voice disorder in surgery treatment,outcome assessment, and in follow-up assessment.

Acknowledgments

The authors thank Drs. Lei Song and Lingeng Lu for their crit-ical review of the article.

This study was supported by the National Natural ScienceFoundation of China (Grant No: 81170925).

Journal of Voice, Vol. 28, No. 6, 2014774

REFERENCES1. Jacobson BH, Johnson A, Grywalsi C, et al. The voice Handicap Index

(VHI):development and validation [J].AmJSpeechLangPathol. 1997;6:66.

2. Lam PK, Chen KM, HoWK, et al. Cross- cultural adaptation and validation

of the Chinese voice handicap index-10 [J]. Laryngoscope. 2006;116:1192.

3. HsiungMW, Lu P, Kang BH, et al. Measurement and validation of the voice

handicap index in voice-disordered patients in Taiwan [J]. J Laryngol Otol.

2003;117:478.

4. Xu W, Han D, Li H, et al. Application of the Mandarin Chinese version of

the Voice Handicap Index. J Voice. 2010;24:702–707.

5. Hirano M. Clinical examination of the voice. New York: Springer Verlag;

1981:81–84.

6. Kempster GB, Gerratt BR, Verdolini Abbott K, et al. Consensus auditory-

perceptual evaluation of voice: development of a standardized clinical pro-

tocol. Am J Speech Lang Pathol. 2009;18:124–132.

7. Wuyt s FL, Bodt MS, Heyning PH. Is the reliability of a visual analog scale

higher than an ordinal scale ? An experiment with the GRBAS scale for the

perceptual evaluation of dysphonia [J]. J Voice. 1999;13:508.

8. Huang ZM, Minifie FD, Kasuya H, et al. Measure of vocal function during

changes in vocal effort level [J]. J Voice. 1995;9:429–438.

9. Katsuhide I, Aliaa AK, Charles NF, et al. Correlation between vocal func-

tions and glottal measurements in patients with unilateral vocal fold paral-

ysis [J]. Laryngoscope. 1997;107:782.

10. Antoine G, Joana R, Jean-Michel T. Objective aerodynamic and acoustic

measurement of voice improvement after phonosurgery. Laryngoscope.

1999;109:656.

11. Wheeler KM, Collins SP, Sapienza CM. The relationship between VHI

scores and specific acoustic measures of mildly disordered voice produc-

tion [J]. J Voice. 2006;20:308.

12. Guimaraes I, Abberton E. An investigation of the voice handicap index with

speakers of Portuguese: preliminary data [J]. J Voice. 2004;18:71.

13. Zur KB, Cotton S, Kelchner L, et al. Pediatric voice handicap index: a new

tool for evaluating pediatric dysphonia [J]. Int J Pediatr Otorhinolaryngol.

2007;71:77.

14. Wuyts FL, De Bodt MS, Molenberghs G, et al. The dysphonia severity

index: an objective measure of vocal quality based on a multiparameter

approach. J Speech Hear Res. 2000;43:796.

15. Katsuhide I, Aliaa AK, Charles NF, et al. Correlation between vocal

functions and glottal measurements in patients with unilateral vocal fold

paralysis. Laryngoscope. 1997;107:782.