Vocal Fold Phase Asymmetries in Patients with Voice Disorders: A Study Across Visualization Techniques

Abstract

Purpose:

The purpose of this study was to examine differences in vocal fold vibratory phase asymmetry judged from stroboscopy, high-speed videoendoscopy (HSV), and the HSV-derived playbacks of mucosal wave kymography, digital kymography, and a static medial digital kymography image of persons with hypofunctional and hyperfunctional voice disorders. Differences between the methods of visual judgments and objective measures of left-right phase asymmetry were assessed. The findings were compared with those from a previous study with vocally-normal speakers.

Methods:

Forty-nine persons with voice disorders underwent stroboscopy and HSV. The HSV images were processed resulting in four different spatial or kymographic displays. Two types of phase asymmetries, left-right and anterior-posterior, were visually rated. Objective measures of left-right phase asymmetry were obtained.

Results:

From stroboscopy, the HSV playback, and the HSV-derived playbacks left-right phase symmetry was judged to be symmetrical in 41%, 32%, and 19% of cases. This difference in playbacks was not seen for anterior-posterior asymmetry. Correlation between visual judgments and objective measures was mild for stroboscopy and moderate-to-high for all HSV-based playbacks.

Conclusions:

The use of kymography appears important for judgments of phase asymmetry. Stroboscopy appears to be sensitive, but possibly not specific, to phase asymmetries. Further development of objective measures is warranted for this feature.

INTRODUCTION

Asymmetry is a prominent component of pathological vocal fold vibration seen in disorders associated with mass or tension changes to the vocal folds (Bless, Hirano, & Feder, 1987; Eysholdt, Rosanowski, & Hoppe, 2003). Overall, vibratory symmetry can be defined as the mirror image movement of the right and left vocal folds. There are several types of vocal fold vibratory asymmetries: left-right phase asymmetries, anterior-posterior phase asymmetries, left-right glottal width asymmetries, frequency asymmetries, and mucosal wave asymmetries. This paper focuses on left-right and anterior-posterior phase asymmetries.

Left-right phase asymmetry is defined as a discrepancy in the phase of the vibratory cycle between the right and left vocal folds. For example, if the right vocal fold begins moving towards midline before the left vocal fold begins to move towards midline, then left-right phase asymmetry would exist. In this study, left-right phase asymmetry is judged and measured as the relationship of the vocal folds at the point of maximal opening during the vibratory cycle.

Anterior-posterior phase asymmetry is defined as a discrepancy in the phase of the cycle between the anterior and posterior aspects of each vocal fold. For example, if the posterior portion of the right vocal fold begins to close before the anterior portion of the right vocal fold, anterior-posterior phase asymmetry would exist. In this study, anterior-posterior phase asymmetry is judged as the relationship of the anterior and posterior aspect of each vocal fold at the point of maximal opening during the vibratory cycle.

Vocal fold vibratory asymmetries are considered diagnostically important (Hirano & Bless, 1993; Stemple, Glaze, & Klaben, 2000). These vibratory asymmetries are believed to be caused by differences in the mass and/or tension of the vocal folds. The mass or tension differences cause the vocal folds to have more or less pliable tissue available for vibration. If the mass or tension of one vocal fold is different than the other vocal fold, the differing physical properties cause the vocal folds to vibrate at slightly different speeds. This is generally not severe enough to cause differences in the frequency of vibration between the two vocal folds, but often presents as phase asymmetry. The information about symmetry of the mass and tension of the vocal folds inferred from assessing left-right phase asymmetry is considered clinically important.

Asymmetry of vocal fold vibration is typically analyzed visually using stroboscopy in the clinic (Bless, Hirano, & Feder, 1987). Stroboscopy provides important clinical information upon which to base decisions about the diagnosis and treatment of persons with voice disorders. Despite its broad clinical use, stroboscopy has known limitations such as relatively low frame rate (30 frames per second (fps)), reliance on pitch tracking and the inability to sample multiple points within one glottal cycle (Eysholdt, Tigges, Wittenberg, & Proschel, 1996; Deliyski, Petrushev, Bonilha, Gerlach, Martin-Harris, & Hillman, 2008). Due to these technical limitations, the asymmetry must be stable over several periods in order to be accurately analyzed using stroboscopy. It is possible that important information about asymmetry is being lost when using stroboscopy alone. This possibility is due, in part, to stroboscopy’s reliance on pitch tracking and the susceptibility of pitch tracking errors to numerous factors in the acoustic signal which would cause the vocal folds to appear out-of-sync. Because HSV does not have similar limitations, the use of stroboscopy with HSV may provide better diagnostic information. HSV records vocal fold vibration with temporal resolution between 2,000 and 10,000 frames per second (fps) (Deliyski, Petrushev, Bonilha, Gerlach, Martin-Harris, & Hillman, 2008). For a voice with a fundamental frequency of 100 Hz, the speed of 2,000 fps provides 20 images for each glottal cycle (Shaw & Deliyski, 2008). While HSV provides intracycle information that is not available from stroboscopy, it relies on the clinician’s judgment and memory over many cycles to make judgments about cycle to cycle variability. The use of kymography derived from HSV recordings overcomes this limitation by allowing the clinician to view one pixel line along the margin of the right and left vocal folds from multiple cycles of vibration combined into one image (Figure 1) (Bonilha & Deliyski, 2008; Švec, Šram, & Schutte, 2007; Wittenberg, Tigges, Mergell, & Eysholdt, 2000).

Figure 1:

Posterior, medial, and anterior position frames from a digital kymography (DKG) playback.

Kymography gives the clinician the ability to analyze the symmetry of many cycles of vibration simultaneously and compare them. Additionally, kymography provides an easy means to obtain objective measurements of vocal fold vibratory features (Eysholdt, Rosanowski, & Hoppe, 2003; Deliyski, Petrushev, Bonilha, Gerlach, Martin-Harris, & Hillman, 2008; Qiu, Schutte, Gu, & Yu, 2003). Given our knowledge of the technological limitations of stroboscopy, it is a natural assumption that HSV, and especially HSV-derived kymography, would be more sensitive to vibratory asymmetry than stroboscopy.

HSV-derived kymography can take several forms. One way it can be viewed is as a playback from the posterior to the anterior of the vocal folds (digital kymography [DKG] playback) (Figure 1). The DKG playback affords the clinician the ability to assess the symmetry of vibration in one image over multiple cycles and over the entire length of the vocal folds. Specifically, the DKG playback displays a kymogram of one pixel line (right to left) of the vocal folds over several cycles which plays as a movie with the pixel line first located in the posterior of the vocal folds and continually scanning anteriorly. This creates a DKG movie where the kymogram shows each line of the vocal folds over time from the anterior to the posterior. Another way that HSV-derived kymography can be viewed is as a static image or as a snapshot of the DKG playback, usually from the middle portion of the vocal folds (medial DKG image) (see the middle image of Figure 1). This method allows the clinician to assess the symmetry of the medial portion of the vocal folds (striking zone) without the interference of the anterior or posterior vocal fold information. This image would be identical to pausing the DKG playback at the medial pixel line. Lastly, it can be derived from a HSV recording that is processed to highlight a specific feature, such as mucosal wave, and played in movie format like the DKG playback (mucosal wave kymography (MKG) (Figure 2). The use of the MKG playback for addressing asymmetry may be important as it highlights a feature that can be used to assess the direction of the vocal fold movement. All three of these HSV-derived kymography forms have been used in this study.

Figure 2:

A medial position frame of a mucosal wave kymography (MKG) playback. Color indicates the phase of glottal edge motion (opening = green, closing = red). See text for more detail.

The assumption that HSV and kymographic and non-kymographic HSV-derived playbacks are more sensitive in the detection of vocal fold asymmetry was supported in research findings from our team on phase asymmetries in normophonic speakers. In this study of 52 vocally-normal speakers, instances of asymmetry were noted more frequently on HSV and HSV-derived playbacks than with stroboscopy (Bonilha, Gerlach, & Deliyski, 2008). The majority of subjects demonstrated left-right and anterior-posterior phase asymmetries. When comparing habitual and pressed phonations in these subjects, pressed phonation was associated with less and more mild asymmetry. Asymmetry was statistically significantly more likely to be present when judged from HSV-derived playbacks than from stroboscopy.

The current study used stroboscopy, HSV and HSV-derived playbacks to assess phase asymmetries in persons with voice disorders. This paper sought to replicate the methodology applied to our previous study of vocally-normal speakers, and as such limited its scope to left-right and anterior-posterior vocal fold vibratory phase asymmetries.

PURPOSE

The purpose of this research was to investigate the variation and magnitude of left-right and anterior-posterior phase asymmetry in persons with voice disorders, and to compare this to the results of vocally-normal speakers. The specific research questions were:

1. What is the rate of occurrence and magnitude of left-right and anterior-posterior phase asymmetry in persons with voice disorders? Furthermore, is the incidence and magnitude of phase asymmetry different in persons with hypofunctional voice disorders compared to persons with hyperfunctional voice disorders?

2. How does the phase asymmetry in persons with voice disorders and vocally-normal speakers compare?

3. How do the rate of occurrence and magnitude of phase asymmetry compare across visualization techniques: stroboscopy, HSV playback, digital kymography (DKG) playback, mucosal wave kymography (MKG) playback, and medial digital kymography (mDKG) images?

4. How do the visually assessed and objectively measured values of left-right phase asymmetry compare?

MATERIALS AND METHODS

Participants

Individuals exhibiting vocal pathology were recruited from voice evaluation referrals to Presbyterian Voice Clinic in Charlotte, NC via verbal request from the research team. Participants were asked to be part of the study upon arriving at the voice clinic for evaluation. They were recruited consecutively without exclusionary criteria except the inability to complete the research procedures. The 49 participants consisted of 39 females and 10 males ranging in age from 18–88 with a mean age of 54 years. Participants in the study signed an informed consent form. The data for this study was recorded at Presbyterian Hospital’s specialized voice center in Charlotte, NC. The speech-language pathologists involved with data collection were specifically trained in voice and the use of HSV.

Visualization of the larynx with rigid endoscopy was used as a means to visually examine the vocal folds and obtain crucial information regarding the nature of the participant’s pathology, as well as how it affected the vocal fold vibratory features. Based on the findings of the examination, the participants were grouped into three categories: hypofunctional (n=21) and hyperfunctional (n=28). For the purposes of this study, a hypofunctional voice was defined as a disorder that results in decreased glottal closure due to muscle weakness or neurological deficit. The hypofunctional group was characterized by glottal insufficiency and included paresis/paralysis and vocal fold atrophy/bowing. For the purposes of this study, a hyperfunctional voice disorder was defined as a voice disorder that results in increased laryngeal and/or supraglottal tension, irrespective of the presence of an organic lesion. The hyperfunctional group was characterized by supraglottal compression and included patients with muscle tension dysphonia and contact lesions. The specific diagnoses of each of the participants can be found in Table 1.

Table 1:

Type of voice disorder and number of participants with each disorder for all persons included in this study.

	Type of Disorder	# of cases
Hypofunctional	Vocal fold atrophy	11
	Vocal fold bowing	3
	Unilateral paralysis	2
	Unilateral weakness	2
	Bilateral paresis	1
	Unilateral paralysis post-arytenoid relocation	1
	Scar	1
Hyperfunctional	Nodules	7
	Post-lesion removal	6
	Generalized edema	5
	Polyp (with or without reactive nodule)	4
	Hemorrhage	2
	Cyst with reactive nodule	2
	Leukoplakia	1
	Muscle tension dysphonia	1
	Total # cases	49

A previously collected database of stroboscopy and HSV recordings from vocally-normal speakers was used as a comparison group. The normophonic speaker database is comprised of twenty-four males and twenty-eight females for a total of 52 participants. Vocal normality was determined by case history, perceptual voice evaluation, voice self-evaluation, and a lack of apparent pathology upon stroboscopy. The database of normophonic speakers has been used in several previous studies (Deliyski, Petrushev, Bonilha, Gerlach, Martin-Harris, & Hillman, 2008; Shaw & Deliyski, 2008; Bonilha & Deliyski, 2008) including a study of phase asymmetry (Bonilha, Gerlach, & Deliyski, 2008). The data collection, storage, and use were in accordance with human subjects regulations approved by the University of South Carolina institutional review board.

Instrumentation and Procedures

Endoscopy and Stroboscopy:

Standard clinical procedures were utilized for rigid endoscopy and stroboscopy. Positioning of the endoscope to place the image of the vocal folds in full view was accomplished with continuous light. Stroboscopy was used to capture one sample of phonation at habitual pitch and intensity levels. A Digital Rhino-Laryngeal Stroboscopic System Model 9100B (Kay Elemetrics Corp., Lincoln Park, NJ) coupled to a 70-degree rigid endoscope (Kay Elemetrics Model 9106) was used along with a laryngeal contact microphone to track vocal fold vibratory frequency. Quiet rooms typically employed for assessment of voice clients in the hospital clinic were the setting for the recordings.

High-Speed Videoendoscopy:

Kay Elemetrics High-Speed Video System Model 9700 equipped with a camera that captures at 2,000 fps for 2.2 seconds with 120×256 pixel resolution was used. A 70-degree rigid endoscope, the same as that used in stroboscopy, and a 300 W constant xenon light source (Kay Elemetrics Model 7152) were coupled with the system. Since high-speed cameras require an intense light source for visualization of the vocal folds, the duration of light exposure was kept at a minimum. The recording of HSV was synchronized with the acoustic recording, captured via a head-mount condenser microphone (AKG C420; AKG Acoustics Harman Pro GmbH, Munich, Germany) to allow for comparisons between physical and acoustic events. Participants were instructed to produce one sample of /i/ at habitual pitch and intensity. Recordings were excluded and the sample was re-elicited if any of the following errors occurred: the intended phonation type was different than that elicited, the view of the vocal folds was insufficient, or the view of the vocal folds was either tilted or had possible motion artifacts that may hamper analysis of the recording.

Image Processing:

The selection of the sample of the HSV recording to be processed and then evaluated was accomplished by using the following criteria: 1) including phonation, 2) excluding onsets, offsets and other transitional events, 3) measuring the glottal area waveform and selecting 100ms before and after the minimum variation of the fundamental frequency to ensure that the frames were representative of the phonation. This was accomplished using the automatic temporal segmentation (based on variability) software detailed in Deliyski, Petrushev, Bonilha, Gerlach, Martin-Harris, & Hillman (2008). Image pre-processing of the HSV recordings included motion compensation and removal of reflection spots. The motion compensation techniques ensure that anatomical structures subjected to kymography are time-aligned. It has been noted that if endoscope motion is unaccounted for, it may affect the validity of the data (Deliyski, 2005). Further, advanced image processing techniques were utilized to convert the enhanced HSV playback into two additional facilitative movies: the DKG playback (Figure 1) and the MKG playback (Figure 2) (Deliyski, Petrushev, Bonilha, Gerlach, Martin-Harris, & Hillman, 2008). From the DKG playback, a static image of a medial line scan was created and termed mDKG.

Visual Perceptual Judgments:

From HSV, three playbacks and one image were rated: the HSV playback, the DKG playback, the MKG playback, and the mDKG static image. The HSV playback provides a visualization of actual vocal fold vibration in slow motion. The DKG playback scans the vocal folds from anterior to posterior producing kymograms of each horizontal line that is viewable in a movie format (Figure 1). The mDKG is a static image of a medial line scan of the DKG playback (middle kymogram of Figure 1). The MKG playback is similar to DKG in allowing for the temporal visualization of the dynamics of the vocal fold edges during glottal opening and closing in consecutive glottal cycles of sustained phonation. In MKG, the color shows the phase of motion of the glottal edges (opening is green and closing is red). The mucosal wave extent appears as a double-edged or thicker curve during the closing phase.

The playbacks obtained from the 49 participants were visually evaluated and rated for phase asymmetry by two clinicians. Playbacks and images from 49 participants using five different viewing modalities amounted to 245 playbacks and images that were judged. In addition, 20% of the playbacks and images were randomly repeated to obtain intra-rater reliability. Therefore, a total of 294 playbacks and images were judged for features of phase asymmetry. Based on a sample size estimation and two raters for forty-nine subjects with an expected correlation coefficient of 0.80 one can expect 95% confidence intervals with half-widths <0.10 (Giraudeau & Mary, 2001; Karanicolas, Bhandari, Moroni, Richardson, Walter, Norman, & Guyatt, 2009). Half-widths of a confidence interval for a particular statistic describe the statistic’s margin of error. Similar to confidence intervals, half-widths are calculated based on a confidence level. The <0.10 value relates to the confidence level selected (greater than 90%) and refers to the range of correlation coefficients. Left-right and anterior-posterior asymmetry were rated on a five point scale as follows: 1= completely asymmetrical, 2= severely asymmetrical, 3=moderately asymmetrical, 4= mildly asymmetrical and 5= symmetrical. This scale is based on common clinical rating methods. Visual perceptual ratings were recorded in a modified version of the ALVIN program (Hillenbrand & Gayvert, 2005), as shown in Figure 3. Information for rating left-right and anterior-posterior asymmetry at one point in time are available in one image from the stroboscopy and HSV. However, judging multiple cycles from one image of stroboscopy and HSV require the rater to hold the percept of extent of asymmetry in memory while watching and evaluating other images. Kymography allows raters to access the symmetry over a number of cycles in one image, but it restricts the image to one pixel line across the vocal folds. Thus, multiple images must be viewed from kymography to assess anterior-posterior asymmetry. Multiple images were viewed in a movie/playback method in MKG and DKG playbacks for the assessment of anterior-posterior asymmetry. Due to the novelty of judging anterior-posterior asymmetry of vocal fold vibration, the overall anterior-posterior asymmetry was reported instead of separate results for the right and left vocal fold. Thus, overall anterior-posterior asymmetry was calculated by taking the average of the ratings for the right and left vocal folds. The difficulty of judging this feature comes from both a lack of familiarity with the feature and a lack of knowledge regarding how common anterior-posterior phase asymmetries are in persons with voice disorders.

Figure 3:

Example of left-right and anterior-posterior phase asymmetry visual-perceptual judgments of DKG playback made via the ALVIN program.

Only left-right and not anterior-posterior phase asymmetry was rated from the mDKG static images. In addition to the visual-perceptual judgments, left-right phase asymmetry was quantified from the mDKG static image. Vocal fold left-right relative phase asymmetry, A (%), was measured over three cycles by taking the time differentials Δ_i between the onset of the closing phase for the right and left vocal folds and dividing the difference by the cycle periods T_i, as seen in Figure 4. The mean A of the three cycles was used to improve measurement precision. Dividing by the cycle periods allowed for normalizing for pitch and camera speed. mDKG images with frequency differences between the vibration of the left and right vocal fold in the mDKG image were excluded from objective measures because left-right phase asymmetry measurements require a clearly defined cycle of vibration.

Figure 4:

Left-right relative asymmetry, A (%), measurement.

Analysis

To compare HSV, DKG, mDKG, MKG, and stroboscopy, the number of symmetrical and asymmetrical judgments and the degree of asymmetry, as previously described, for each rater was reported. A Pearson Correlation statistic was calculated to determine the relationship between the objective measurements made from the mDKG and subjective ratings recorded for the two observers. Correlation was considered low, mild, moderate and high for r < 0.25, 0.25 ≤ r < 0.5, 0.5 ≤ r < 0.75 and r ≥ 0.75, respectively. Inter- and intra-rater reliability was established by evaluation of the percent agreement between the measures. Percent agreement was considered acceptable at 75%. This threshold was selected based on a review of the literature, specifically Poburka (1999) and Rosen (2005). Fisher’s Exact Test was employed to determine statistical significance (p=0.05) between levels of each asymmetry and between the viewing modalities. Due to the preliminary nature of this study and the difference expected in each of the viewing modalities, an alpha level of 0.05 was considered sufficiently rigorous.

RESULTS

Left-Right Phase Asymmetry

Left-right phase asymmetry (Table 2) was observed in the majority of cases irrespective of the type of playback used for evaluation. The majority of asymmetry identified was considered to be mild. Left-right phase asymmetries were identified most frequently on the MKG playbacks (93%). MKG-based symmetrical ratings (6) were statistically significantly less frequent than symmetrical ratings from mDKG (30%) (p=0.004) and HSV (41%) (p=0.00). The highest percentage of severely or completely asymmetrical ratings was from the non-kymographic playbacks of HSV (15%) and stroboscopy recordings (12%), while the lowest percentage was from the MKG playback (8%). These differences were not statistically significant (p=0.5 and p=0.37). [Insert Table 2]

Table 2.

Percentage of voice-disordered participants with left-right phase asymmetry for each visualization technique according to two raters.

Technique	Completely Asymmetrical	Severely Asymmetrical	Moderately Asymmetrical	Mildly Asymmetrical	Symmetrical
DKG
Rater 1	4	10	27	49	10
Rater 2	2	6	18	43	29
MKG
Rater 1	4	6	35	51	4
Rater 2	4	2	49	35	10
mDKG
Rater 1	6	2	24	49	18
Rater 2	2	10	16	31	41
HSV
Rater 1	4	8	16	45	27
Rater 2	2	16	6	20	55
Stroboscopy
Rater 1	6	10	12	59	12
Rater 2	2	6	4	37	51

Note. DKG= Digital Kymography; MKG= Mucosal Wave Kymography; mDKG = medial line of Digital Kymography; HSV = high-speed videoendoscopy

Anterior-Posterior Asymmetry

Anterior-posterior asymmetry was noted for the large majority of persons across playbacks as displayed in Table 3. Anterior-posterior phase asymmetries were slightly less likely to be noted on stroboscopy. The majority of anterior-posterior asymmetries were judged as mild. The greatest number of “severe” anterior-posterior phase asymmetry ratings was from analysis of the DKG playback (p=0.002). This resulted in statistically significantly fewer instances of mild ratings from the DKG playback compared to MKG (p=0.0017) and stroboscopy (p=0.033).

Table 3.

Percentage of voice-disordered participants with anterior-posterior phase asymmetry for each visualization technique according to two raters.

Technique	Completely Asymmetrical	Severely Asymmetrical	Moderately Asymmetrical	Mildly Asymmetrical	Symmetrical
DKG
Rater 1	2	17	39	42	0
Rater 2	2	6	45	47	0
MKG
Rater 1	2	0	35	63	0
Rater 2	4	0	22	71	2
HSV
Rater 1	2	2	38	56	2
Rater 2	2	14	49	35	0
Stroboscopy
Rater 1	2	3	34	56	5
Rater 2	1	4	57	33	5

Note. Abbreviations same as Table 2.

Persons with Voice Disorders vs. Vocally-Normal Speakers

There were more reported cases of left-right phase asymmetry that were considered moderate to complete, identified in patients with voice disorders than in vocally-normal speakers (Table 4). This difference was statistically significant for stroboscopy, DKG, and MKG at p=0.03, 0.01, and 0.00 respectively. Analysis of stroboscopy and HSV revealed fewer reported cases considered to demonstrate severe asymmetry, although this difference was slight in comparison to mDKG and DKG. A difference in the rate of occurrence of larger asymmetry was found to a greater extent for anterior-posterior phase asymmetry where a statistically significant difference between vocally-normal individuals and patients with voice disorders was found for ratings from all playbacks at p=0.02-p=0.00 (Table 5).

Table 4.

Percentage of voice-disordered participants judged for left-right phase asymmetry based on each visualization technique.

Technique	> Mildly Asymmetrical			Mildly Asymmetrical			Symmetrical
	Disorder	Normal#	p	Disorder	Normal#	p	Disorder	Normal#	p
DKG	35	14	0.01*	46	58	0.19	19	28	0.16
MKG	50	10	0.00*	43	52	0.24	7	38	0.00*
mDKG	31	18	0.09	40	50	0.24	29	32	0.41
HSV	26	19	0.26	33	34	0.50	41	47	0.37
Stroboscopy	20	6	0.03*	48	24	0.18	32	70	0.00*

Notes. Abbreviations same as Table 2.

^#Vocally normal control subjects from Bonilha et al. (2008)

^*p < 0.05

Table 5.

Percentage of voice-disordered participants judged for anterior-posterior phase asymmetry based on each visualization technique.

Technique	> Mildly Asymmetrical			Mildly Asymmetrical			Symmetrical
	Disorder	Normal#	p	Disorder	Normal#	p	Disorder	Normal#	p
DKG	56	2	0.00*	44	71	0.00*	0	27	0.00*
MKG	32	1	0.00*	67	70	0.5	1	29	0.00*
HSV	54	18	0.00*	45	59	0.1	1	23	0.00*
Stroboscopy	51	1	0.00*	44	65	0.02*	5	34	0.00*

Notes. Abbreviations same as Table 2.

^#Vocally normal control subjects from Bonilha et al. (2008)

^*p < 0.05

Hypofunctional vs. Hyperfunctional Voice Disorders

The reported cases of hypofunctional voice disorder had a lower rate of occurrence of left-right phase asymmetry than cases of hyperfunctional voice disorder as seen in the bottom row of Table 6 (symmetrical vs. asymmetrical). This difference was statistically significant for the mDKG images. There was not a clear difference between hypofunctional and hyperfunctional voice disorders in the rate of occurrence of anterior-posterior phase asymmetries (Table 7).

Table 6.

Percentage of hypofunctional and hyperfunctional voice-disordered participants judged for left-right phase asymmetry based on each visualization technique.

	Completely Asymmetrical			Severely Asymmetrical			Moderately Asymmetrical			Mildly Asymmetrical			Symmetrical
	Hypo	Hyper	p	Hypo	Hyper	p	Hypo	Hyper	p	Hypo	Hyper	p	Hypo	Hyper	p
DKG	0	4	1	5	7	1	29	22	0.74	33	41	0.56	33	26	0.54
MKG	5	4	1	5	0	0.43	29	37	0.76	48	56	0.57	14	4	0.30
mDKG	0	4	1	5	7	1	24	19	0.73	10	48	0.004*	62	22	0.007*
HSV	5	0	0.43	5	11	0.63	19	11	0.44	5	37	0.014*	67	41	0.15
Strobe	5	0	0.43	10	4	0.57	10	4	0.57	33	48	0.38	43	44	1

Notes. DKG= Digital Kymography; MKG= Mucosal Wave Kymography; mDKG = medial line of Digital Kymography; HSV = High-Speed Videoendoscopy; Strobe = Stroboscopy

^*p < 0.05

Table 7.

Percentage of hypofunctional and hyperfunctional voice-disordered participants judged for anterior-posterior phase asymmetry based on each visualization technique.

	Completely Asymmetrical			Severely Asymmetrical			Moderately Asymmetrical			Mildly Asymmetrical			Symmetrical
	Hypo	Hyper	p	Hypo	Hyper	p	Hypo	Hyper	p	Hypo	Hyper	p	Hypo	Hyper	p
DKG	0	4	1	0	7	0.50	38	33	0.77	62	56	0.78	0	0	1
MKG	0	4	1	0	0	1	24	19	0.73	71	78	0.74	5	0	0.43
HSV	5	0	0.43	0	0	1	33	44	0.56	62	56	0.78	0	0	1
Strobe	0	0	1	10	0	0.18	19	44	0.12	62	56	0.78	10	0	0.18

Notes. Abbreviations same as Table 6.

^*p < 0.05

Inter- and Intra-rater Reliability

High inter- and intra-rater reliability within one scalar level was established for judges across the five playbacks. If rater 1 judged the asymmetry to be complete and rater 2 judged the asymmetry to be severe for 10 ratings, they would have 100% agreement within one scalar level. When the intra-rater reliability data from the two raters was considered together, agreement was from 82 to 100% within one scalar level, with a mean of 97%. Intra-rater reliability for rater 1 alone, ranged from 91–100% with a mean of 99%. For rater 2, alone, intra-rater reliability ranged from 82–100% with a mean of 94%. Inter-rater reliability for asymmetry ranged from 90 to 100% agreement within one scalar level, with a mean of 96%. The differences in rating asymmetry seemed to be directly related to both the rater’s experience rating vocal fold vibration with HSV, and clinically using vocal fold visualization. That is, the rater with more overall experience with laryngeal endoscopy and more specific experience with HSV had better intra-rater reliability.

Comparison of Playback Modalities

Statistically significant differences in ratings of left-right phase asymmetries by Rater 1 were found for ratings of symmetrical between DKG and HSV(p=0.03) and MKG and HSV(p=0.002), and for moderate between MKG and HSV(p=0.03) and MKG and stroboscopy(p=0.01). For Rater 2, statistically significant differences in ratings of left-right phase asymmetries were found for ratings of symmetrical between DKG and MKG (p=0.020) and DKG and HSV (p=0.01), for moderate between MKG and mDKG (p=0.00) and mDKG and stroboscopy (p=0.046), and for severe between MKG and HSV (p=0.02). Statistically significant differences in ratings of anterior-posterior phase asymmetries by Rater 1 were found for ratings of mild between DKG and MKG (p=0.03) and for severe between DKG and MKG (p=0.00), DKG and HSV (p=0.02), and DKG and stroboscopy (p=0.05). For Rater 2, statistically significant differences in ratings of anterior-posterior phase asymmetries were found for ratings of mild between DKG and MKG (p=0.01), MKG and HSV (p=0.00), and stroboscopy and MKG (p=0.00) and for moderate between MKG and stroboscopy (p=0.00).

Pearson’s correlation was used to compare the rating results across modalities for left-right phase asymmetries (Table 8) and anterior-posterior phase asymmetries (Table 9). Given n=45 and p=0.05, a correlation coefficient above 0.288 signifies a statistically significant relationship between modalities, and a correlation coefficient below 0.288 fails to find a statistically significant relationship between modalities. For left-right phase asymmetry, a lack of a statistically significant relationship between modalities was found for comparisons of: stroboscopy with mDKG, stroboscopy with DKG, and stroboscopy with MKG. The comparisons of mDKG and DKG were the only relationships with a high correlation coefficient for left-right phase asymmetry. Moderate correlation coefficients between modalities for rating left-right phase asymmetry were found for comparisons of HSV and mDKG, HSV and MKG, mDKG and DKG, and DKG and MKG. For anterior-posterior phase asymmetry, statistically significant relationships between modalities were found for comparisons of all modalities for at least one rater. No comparisons had a high correlation coefficient for anterior-posterior phase asymmetry. Moderate correlation coefficients between modalities for rating anterior-posterior phase asymmetry were found for comparisons of DKG and MKG. All other correlation coefficients were mild or low.

Table 8:

Correlation (Pearson) across playbacks for left-right phase asymmetry.

	Rater 1	Rater 2
Strobe & HSV	0.41	0.38
Strobe & mDKG	0.01	0.15
Strobe & DKG	0.18	0.02
Strobe & MKG	0.28	0.03
HSV & mDKG	0.72	0.61
HSV & DKG	0.75	0.69
HSV & MKG	0.73	0.59
mDKG & DKG	0.78	0.80
mDKG & MKG	0.42	0.37
DKG & MKG	0.67	0.71

Note. Abbreviations same as Table 6.

Table 9:

Correlation (Pearson) across playbacks for anterior-posterior phase asymmetry.

	Rater 1	Rater 2
Strobe & HSV	0.43	0.20
Strobe & DKG	0.32	0.17
Strobe & MKG	0.30	0.21
HSV & DKG	0.29	0.40
HSV & MKG	0.19	0.37
DKG & MKG	0.65	0.58

Note. Abbreviations same as Table 6.

Objective Measures of Left-Right Asymmetry

Objective measures were separated into two groups, for hypofunctional (n=21), and for hyperfunctional voice disorders (n=25). One person with a hyperfunctional voice disorder was excluded due to left-right vocal fold vibratory frequency differences. Two recordings were excluded due to poor image quality. The majority of speakers demonstrated measurable left-right relative phase asymmetry ranging from 0 to 10% as seen in Figure 5. Furthermore, the majority of persons with hypofunctional voice disorders had less asymmetry (less than 8%) when compared to persons with hyperfunctional voice disorders (less than 10%). The majority of persons without voice disorders had even less asymmetry (less than 6%). When comparing the groups based on asymmetries that were 6% or less in magnitude, there were 34%, 24%, 54%, and 66% of cases from persons with hypofunctional voice disorders, persons with hyperfunctional voice disorders, and vocally-normal speakers producing habitual and pressed phonations respectively. There were statistically significantly smaller asymmetries in the persons without voice disorders than the persons with voice disorders. The mean difference in magnitude of asymmetry between hyperfunctional and hypofunctional voice disorders as tested by Fisher’s Exact test was not statistically significant at p=0.05.

Figure 5:

Objective measures of left-right phase asymmetry for persons with hypofunctional and hyperfunctional voice disorders and vocally-normal speakers producing habitual and pressed phonations.

The correlation between objective measures of left-right phase asymmetry and visual ratings of left-right asymmetry were mild for stroboscopy (0.39), moderate for HSV (0.58), MKG (0.71) and DKG (0.74), and high for and mDKG (0.80) (compared to means of the 2 raters). In comparison, the previous normophonic study found correlations of 0.19 for stroboscopy, 0.46 for HSV, 0.56 for MKG, 0.54 for DKG, and 0.58 for mDKG (means of the 3 raters) (Bonilha, Gerlach, & Deliyski, 2008). These correlations are higher than those found in the study of vocally-normal speakers presumably due to a wider range of asymmetry in the voice disordered population and increased training prior to visual judgments.

DISCUSSION

Visual judgments

This study sought to determine the presence and magnitude of left-right and anterior-posterior phase asymmetry of vocal fold vibration in persons with voice disorders and determine whether these findings differed by: presence of a voice disorders, type of voice disorder and playback modalities. Left-right and anterior-posterior phase asymmetry were both present in the majority of persons with voice disorders. In general, there was a higher rate of occurrence of anterior-posterior phase asymmetry than left-right phase asymmetry and the majority of asymmetries were mild, patterns that are similar to what has been found in vocally-normal speakers. Judgments of asymmetry magnitude differed more than presence based on voice disorder status and playback modality. These findings speak to the usefulness of phase asymmetry as a diagnostic feature of vocal fold vibration and the importance of understanding the different ways it is viewed from various playback modalities.

Stroboscopy, DKG, and MKG were more sensitive in identifying asymmetry than HSV and mDKG. This finding agrees with data from Kendall (2009) who also found stroboscopy to appear more sensitive to left-right phase asymmetries than the HSV playback. Despite the concordance with Kendall (2009), this finding does not agree with our knowledge of the technical limitations of stroboscopy. Therefore, we completed an in-depth review of the stroboscopy recordings and ratings in the present study. Through this review of individual differences, it became clear that the majority of reported cases of asymmetry were noted to correspond to cases of vocal fold vibratory atypicalities. This finding underscores the technological limitations of stroboscopy and the potential technological benefits of HSV (Deliyski, Petrushev, Bonilha, Gerlach, Martin-Harris, & Hillman, 2008) including HSV’s non-reliance on pitch tracking and the temporal resolution of the recordings.

Additional proof of sensitive, yet not specific, results from stroboscopy can be found in the results from the objective measures and the comparison of results from persons with and without voice disorders. There were 14 cases of symmetry from the objective measures, while there were 13 reported cases of symmetrical ratings from stroboscopy. This finding suggests that stroboscopy is as able to reveal vocal fold vibratory left-right phase asymmetry as well as kymography with a temporal resolution of 8,000 fps. It is more likely that vibratory irregularities or other vibratory atypicalities are responsible for the large number of reported cases (87%) of asymmetry in stroboscopy. This reasoning is especially true as symmetry variation can cause irregularity of the acoustic signal used for stroboscopy synchronization (Mehta, Deliyski, & Hillman, 2010; Mehta, Deliyski, Zeitels, Quatieri, & Hillman, 2010). Further support that stroboscopy’s asymmetry sensitivity is false comes from our findings in vocally-normal speakers. Stroboscopy detected less asymmetry than the other imaging modalities in the prior normative study (Bonilha, Gerlach, & Deliyski, 2008). If stroboscopy were truly more sensitive to left-right asymmetry one would expect this sensitivity to be seen in both studies of persons with and without voice disorders. It is likely that the visual perceptual judgments of asymmetry are affected or biased by features other than left-right phase asymmetry that are more predominant in the endoscopic recordings of persons with voice disorders compared with persons without voice disorders. Given these evidences, one would conclude that stroboscopy is neither more sensitive nor more specific to left-right phase asymmetries than HSV or HSV-derived playbacks. This conclusion has clinical implications that stress the need for thoughtful interpretation of vocal fold vibratory recordings when deciding whether the asymmetry is real or the result of technical factors or other vibratory irregularities.

Although the majority of phonations were characterized by left-right asymmetry, the asymmetry was generally mild. This result was also found in the vocally-normal study. The HSV-derived playbacks were statistically significantly more sensitive to left-right phase asymmetries than the HSV playback. This finding highlights the importance of kymography and facilitative playbacks from HSV to improve the clinical usefulness of the recordings. The similarities between the rate of occurrence of asymmetry in persons with voice disorders and in normophonic persons stresses the need to use caution when evaluating the normality of asymmetry in patients with voice disorders to prevent misdiagnosis or over-diagnosis. While the frequency of cases reported to demonstrate greater than mild asymmetry was larger in persons with voice disorders than in those without voice disorders, this was only statistically significant for the DKG and MKG playbacks. Additionally, the overlap in asymmetry presence these two groups demonstrates the importance of magnitude of asymmetry in the clinical utility of the visual rating. Both for persons with and without voice disorders, anterior-posterior phase between asymmetries were reported in almost all cases with the majority of cases being mild in severity. This finding again stresses the need for increased knowledge of normal and pathological vocal fold vibration prior to confident use of phase asymmetry as a secure biomarker for voice disorders.

Differences between hypofunctional and hyperfunctional voice disorders were relevant for left-right phase asymmetries. The mechanism of hyperfunctional voice disorders more clearly associates it with left-right vocal fold vibratory phase asymmetry than the mechanism of hypofunctional voice disorders. The majority of the participants in this study who demonstrated hyperfunction also had a vocal fold lesion or notable edema. The differences in the mass between the two vocal folds would be expected to produce asymmetrical vibration. The majority of participants in this study demonstrating hypofunction had bilateral atrophy or bowing, thus they would be expected to have fewer instances higher magnitude of asymmetry given the bilateral nature of the disorder. When asymmetry was present, no large differences were noted in the severity of the asymmetry between these two groups. Thus, it seems that presence of asymmetry rather than magnitude of asymmetry may be the important marker for differentiating voice disorders. However, severity of asymmetry is still important for differentiating normal vibration from pathological vibration.

The finding of a high rate of occurrence of left-right phase asymmetries in persons with hyperfunctional voice disorders is opposite of what was predicted from the results of vocally-normal speakers (Bonilha, Gerlach, & Deliyski, 2008). When vocally-normal speakers produced pressed phonation, there was a reduced rate of occurrence of left-right phase asymmetries. Thus, it was assumed that the compression increased the subglottal air pressure to overcome the asymmetries that were visible in the speaker’s habitual phonation. This same mechanism was thought to be applied, to a lesser extent, as a natural compensation in persons with hyperfunctional voice disorders. Often such patients with voice disorders are seen to overdrive the system to produce a clearer voice. However, evidence from this study suggests that patients with hyperfunctional voice disorders may not use this mechanism, or may not use it to the anticipated degree, to achieve sustained phonation at habitual pitch and loudness levels. Further investigation of the mechanisms of natural compensation for hyperfunctional voice disorders is warranted to discern the need for vocal unloading exercises and repeated endoscopic evaluation prior to final diagnosis. There were no relevant differences for anterior-posterior phase asymmetries between hypofunctional and hyperfunctional voice disorders.

Rater reliability is typically judged within one scalar level. For example, if rater 1 judges asymmetry to be a 2 (severely asymmetrical) and rater 2 judges asymmetry to be a 3 (moderately symmetrical), these ratings would be considered to agree. In this study, average intra- and inter-rater reliability within one scalar level was high at 97% and 96%, respectively. Inter- and intra-reliability are reported in this way for visual-perceptual judgments of vocal fold vibratory features due to the lack of a reliable, valid and standardized method to rate these features. The literature indicates that raters have at best moderate exact reliability when judging vocal fold vibratory features even though these are the precise judgments that are used in the clinic for diagnosis and measuring treatment outcomes. In this study, average exact intra- and inter-rater reliability were 75% and 59%, respectively. These averages are higher than those typically reported but still demonstrate less than adequate reliability for a diagnostic test. There are two general approaches to improving the reliability of information on vocal fold vibratory features: 1) develop a standardized rating system based on experiments that establish the precise levels of each feature that a rater can reliably discern and 2) use more automated, objective measures of vocal fold vibration. While automated, objective measures do not have the inherent errors related to rater reliability, they are removed from the anatomy and the physiology of the disorder making it difficult to base surgical procedure decisions on if used as the sole technique. Thus, there is likely a role for both the objective measures and the possibly more clinically-intuitive ratings to improve information on vocal fold vibratory features.

Objective measures

Objective measures were moderately correlated with subjective ratings of left-right phase asymmetry from HSV recordings demonstrating the importance of further efforts to quantify vocal fold vibratory features from HSV recordings. Correlation between objective measures and visual ratings was lowest for stroboscopy and highest for the HSV-derived kymography playbacks and images (mild for stroboscopy (0.39), moderate for HSV (0.66), and high for MKG (0.76), DKG (0.78), and mDKG (0.82)). The mild correlation between objective measures of asymmetry and visual ratings of asymmetry from stroboscopy provides further evidence that stroboscopy is sensitive, but not specific, to vocal fold vibratory asymmetries. It is likely that the high correlation for mDKG, and to a lesser degree MKG and DKG, is due to the fact that the objective measures were made from the mDKG images and, thus, correlated the best with the ratings from those images. Furthermore, the stroboscopy and HSV recordings were made consecutively, but not simultaneously. Even though these examinations were from the same person recorded within minutes of each other and in a carefully controlled environment, they are different recordings of different phonations and as such increase the variability between the stroboscopy playback and the HSV and HSV-derived modalities. No correlation was above 0.80 emphasizing the limitations of our visual perceptual system for assessing this feature. The correlations between objective measures and visual-perceptual ratings were higher in persons with voice disorders than in vocally-normal speakers across all displays. The difference between the objective measures and visual ratings stress the importance of developing automated objective measures for left-right and anterior–posterior phase asymmetries for both clinical purposes and as a means to study about our visual-perceptual judgments of vocal fold vibratory features.

Clinical Implications

There are three main clinical implications from the results of this study. First, since stroboscopy is the widely used “gold standard” for evaluating vocal fold vibratory behavior, the results of this study that suggest that left-right phase asymmetry is not able to be specifically judged from stroboscopic recordings is directly relevant to clinical practice. It is vitally important to understand the limitations of the technology used to diagnosis voice disorders. Clinicians using stroboscopy in their evaluation should be aware that the results of this study suggest that visible left-right phase asymmetry viewed in stroboscopic recordings may be more indicative of other vocal fold vibratory atypicalities and not precisely reflective of left-right phase asymmetry alone.

Second, in left-right and anterior-posterior phase asymmetries, patients with voice disorders had a greater rate of occurrence of moderate to complete asymmetry ratings than persons without voice disorders. Both groups had large rates of occurrence of mild asymmetry ratings. This finding is clinically relevant because it supports the theory that severity and not presence of phase asymmetry is the key aspect of the variable. This supplements the results of the previous work in phase asymmetry in vocally-normal speakers. The results of these findings applied to clinical cases would suggest that clinicians view mild left-right and anterior-posterior phase asymmetries as within normal limits.

Third, based on the correlations between the objective measures and visual-perceptual ratings of left-right phase asymmetry from HSV and stroboscopy playbacks, ratings from HSV playbacks are more accurate in identifying left-right phase asymmetry than ratings from stroboscopy. While this finding suggests that HSV playbacks are better for judging left-right phase asymmetries there are two caveats to this discrepancy in accuracy. The first caveat is, as mentioned in the limitations section, HSV playbacks and the objective measures were taken from a different endoscopic recording than the stroboscopy playbacks. While different recordings can be the reason for small differences in left-right phase asymmetry, it is unlikely by which the results would be due solely to this given the sequential method that the recordings were attained. The second caveat is that while this study found that ratings from HSV playbacks are more accurate than ratings from stroboscopy playbacks for left-right phase symmetry, this study did not evaluate the clinical relevance of the difference in accuracy between the two techniques. That is, the accuracy difference may be clinically significant and indicate that HSV recordings are preferable to stroboscopy recordings for judgments of left-right phase asymmetry or the accuracy difference may not be clinically significant and indicate that either HSV or stroboscopy recordings are useful for the assessment of left-right phase asymmetry. These same discussions of clinical implications can be applied to the use of HSV playbacks versus the use of the HSV-derived kymographic modalities as the kymographic modalities had higher correlation with the objective measures than the HSV playback.

Limitations

There are three caveats to this study that should be taken into consideration when reviewing the results. For clinical use, there is no one vocal fold vibratory characteristic that is sufficient to define a vocal pathology. The second caveat is that HSV has not yet been shown through empirical research to provide information which changes a diagnosis or a management plan. While we have had positive anecdotal experience of HSV influencing both diagnoses and management plans, we remain cautious regarding the impact that HSV will have in the voice clinic until research has substantiated our personal experiences. The third caveat is the division of the data into hypofunctional and hyperfunctional voice disorders. We recognize that this division is not universally accepted, but feel that it is a manageable method for providing further information regarding phase asymmetries and the use of stroboscopy, HSV, and HSV-derived kymography.

CONCLUSIONS

The majority of persons with voice disorders exhibit left-right and anterior-posterior phase asymmetries when rated and measured from videoendoscopic recordings. The majority of these asymmetries are mild; however a few are severe which was not found in vocally-normal speakers. Persons with hypofunctional voice disorders demonstrated fewer asymmetries than persons with hyperfunctional voice disorders both visually and objectively. However, both patient groups displayed more and larger left-right asymmetries via objective measurement than the vocally-normal speakers. Left-right asymmetries were detected more readily via HSV-derived playbacks using kymography than HSV alone. Stroboscopy appears highly sensitive to vocal fold vibratory asymmetry, but is likely sensitive to all variability in vocal fold vibration and pitch and possibly not specifically sensitive to asymmetry. Discrepancies between objective measures and visual-perceptual ratings of left-right asymmetry reveal the necessity for future quantitative analysis techniques to strengthen both research and clinical applications of laryngeal visualization. Future investigations should compare these findings to those from HSV recordings accomplished at higher frame rates to ensure that HSV findings are not a result of technical constraints.