Optical Flow Analysis of Paralaryngeal Muscle Movement

Abstract

Objectives:

The paralaryngeal muscles are thought to be hyperfunctional with phonation in patients with primary muscle tension dysphonia (pMTD). However, objective, quantitative tools to assess paralaryngeal movement patterns lack. The objectives of this study were to (1) validate the use of optical flow to characterize paralaryngeal movement patterns with phonation, (2) characterize phonatory optical flow velocities and variability of the paralaryngeal muscles before and after a vocal load challenge, and (3) compare phonatory optical flow measures to standard laryngoscopic, acoustic, and self-perceptual assessments.

Methods:

Phonatory movement velocities and variability of the paralaryngeal muscles at vocal onsets and offsets were quantified from ultrasound videos and optical flow methods across 42 subjects with and without a diagnosis of pMTD, before and after a vocal load challenge. Severity of laryngoscopic mediolateral supraglottic compression, acoustic perturbation, and ratings of vocal effort and discomfort were also obtained at both time points.

Results:

There were no significant differences in optical flow measures of the paralaryngeal muscles with phonation between patients with pMTD and controls. Patients with pMTD had significantly more supraglottic compression, higher acoustic perturbations, and higher vocal effort and vocal tract discomfort ratings. Vocal load had a significant effect on vocal effort and discomfort but not on supraglottic compression, acoustics, or optical flow measures of the paralaryngeal muscles.

Conclusion:

Optical flow methods can be used to study paralaryngeal muscle movement velocity and variability patterns during vocal productions, although the role of the paralaryngeal in pMTD diagnostics (e.g., vocal hyperfunction) remains suspect.

Lay Summary:

Paralaryngeal muscle movement using optical flow methods and standard clinical metrics were compared between subjects with and without a primary muscle tension dysphonia. There were significant group differences on clinical metrics but not on optical flow measures (265 of 280 max)

INTRODUCTION

Increased or excessive movement in paralaryngeal muscles has long been attributed to vocal hyperfunction and dysfunction in patients with functional voice disorders like “muscle misuse” disorders and primary muscle tension dysphonia (pMTD).^1-3These paralaryngeal muscle groups include the suprahyoid (geniohyoid, mylohyoid, anterior digastric) and infrahyoid (thyrohyoid, sternohyoid, sternothyroid) muscles, with attachments at the hyolaryngeal framework and other parts of the head and neck. The paralaryngeal muscles (also known as “extrinsic laryngeal muscles”—although technically only the thyrohyoid and sternothyroid muscles attach directly to the larynx) stabilize and support the larynx, are active during airway protection (swallowing),⁴ and may be involved in pitch control.^5,6 However, the role of the paralaryngeal muscles for vocal production are not well understood,^7,8 and their contributions to vocal hyperfunction and dysfunction is even less well-supported.^2,9,10 There is a long-standing theory that increased vocal effort, fatigue, and discomfort patients with pMTD commonly experience with voice use is the byproduct of hyperfunction in these paralaryngeal muscles.^11-13 This theory has been further supported by observations of static increases in laryngeal elevation and reduced thyrohyoid space in patients with pMTD, thought to be the result of increased paralaryngeal muscle contraction.^14,15

However, studies on paralaryngeal movement patterns with phonation that can lead to vocal dysfunction in pMTD lack. One major limitation is the gap in objective methods that can directly quantify velocity and variability of paralaryngeal muscle movement with phonation. The most commonly used methods to assess paralaryngeal muscle function include ventral neck palpation ratings and electromyography. However, ventral neck palpation has been shown to have poor sensitivity and specificity.^16,17 Electromyography of the ventral neck provides information on the level of motor unit recruitment in the neck muscles,^18-20 but not on muscle movement patterns (velocity, variability) because the muscles are not visualized with these methods. These shortcomings have likely led to inconclusive findings across studies on the role of the paralaryngeal muscles in pMTD.^10,17,21-23 Methods that are objective, quantifiable, and can directly study biomechanical properties of the these muscles with phonation between individuals with and without pMTD are needed.

One potential method to address these methodological shortcomings is optical flow, which involves the use of computational imaging processing to detect and study object movements, including magnitude, velocity, vector directions, and movement variability of an object. The concept of optical flow works by first taking a video of a moving object and then analyzing the change in individual pixels or clusters of pixels within that object between two consecutive frames to calculate pixel displacement over time (i.e., optical flow velocity). These methods have previously been applied to robots to make their movement patterns more human-like and to monitor and optimize self-driving cars.^24,25 Optical flow methods have also been shown to be highly reliable and accurate for the study of muscle contraction biomechanics in leg muscles/tendons,^26-29 the heart,³⁰ and even humming bird wings.³¹

Considering patients with pMTD are thought to have paralaryngeal muscle hyperfunction with phonation, it would be expected that patients with pMTD would have greater phonatory optical flow velocities in the paralaryngeal muscles compared to vocally healthy controls. We recently discovered that static contraction (paralaryngeal muscle tension) using shear wave elastography ultrasound methods did not differ between individuals with and without pMTD, suggesting similar motor patterns in the paralaryngeal muscles between groups (at least during at rest conditions and during steady state vowel productions).⁹ However, we did previously observe greater movement variability in the ventral neck (especially near the thyrohyoid space) using motion capture technology during speech tasks.^32,33 These findings suggest there may be group differences in movement patterns within the paralaryngeal muscles. As such, studying paralaryngeal movement velocity and variability could provide additional insights into potential neuromuscular mechanisms underlying vocal dysfunction in patients with pMTD.

The first goal of this study was to apply optical flow methods to ultrasound videos of the suprahyoid and infrahyoid paralaryngeal muscle groups, acquired during phonation, and compare paralaryngeal movement velocity and variability in subjects with and without pMTD. The second goal was to compare phonatory optical flow measures before and after a vocal load challenge in the same groups, as increased vocal demands are a common precursor to pMTD.² The third goal was to compare phonatory optical flow measures to standard acoustic, laryngoscopic, and self-perceptual clinical metrics to study relationships between paralaryngeal muscle movement patterns and standard clinical metrics of vocal dysfunction to better define what is meant by vocal “hyperfunction” of the paralaryngeal muscles.

METHODS

The HIPAA compliant protocol was approved by the University of Texas at Southwestern Medical Center (UTSW) Institutional Review Board. Variables were acquired once before and once after a half-hour vocal load challenge. The study design had two sets of independent variables: group (pMTD, control) and time point (pre-vocal load, post-vocal load) and the dependent variables were optical flow velocity and variability of the (1) suprahyoid paralaryngeal muscles (geniohyoid, mylohyoid, anterior digastric) in the trans-midline view and (2) infrahyoid paralaryngeal muscles (thyrohyoid, sternohyoid, sternothyroid) in the right/left longitudinal view thought to be involved in vocal hyperfunction.^22,23 The secondary dependent variables were self-ratings of vocal effort (100 mm visual analog scale), vocal tract discomfort (Vocal Tract Discomfort Scale^34,35), cepstral peak prominence (CPP) (acoustic voice sample), and mediolateral supraglottic compression (laryngoscopy).

Participants

Forty-two (42) individuals with (n=21) and without (n=21) pMTD participated in the study. Participants with pMTD (mean age: 49.24 years; age range: 23-79 years; 81% women) were recruited at their initial evaluation at the UTSW Center for Voice Care and received a formal diagnosis of pMTD³⁶ at the time of their visit. The diagnosis was made by one of three board-certified laryngologists, who based their diagnosis on case history, laryngoscopic presentation, and palpation. The subjects in the pMTD group were diagnosed based on the Classification Manual of Voice Disorders’³⁶ definition of aberrant laryngeal presentation, which can include the presence of any of the following laryngeal patterns: vocal fold hyperadduction, anterior-posterior supraglottic compression, or mediolateral supraglottic compression. All patients diagnosed with pMTD felt their voice disorder significantly impacted their quality of life (> 11 out of 40 on the Voice Handicap index-10 (VHI-10)³⁷, based on normative values³⁸), and reported pathophysiological levels of vocal fatigue, based on scores more than 24 on Part 1 of the Vocal Fatigue Index (VFI-Part1).³⁹ Acoustic and aerodynamic metrics were not used as inclusion criteria because they have previously been shown to lack sensitivity for the pMTD population,⁴⁰ and because patients with pMTD can report increased vocal effort, fatigue, and discomfort, without overt changes to the sound of their voice.^40-42 We chose vocal fatigue as one of the inclusion criteria symptoms since this parameter is frequently reported in patients with pMTD.² Vocal effort and vocal tract discomfort are also commonly reported in this population but were not used as inclusion criteria since they were included as outcome variables in the experimental design. Although presence of the above mentioned laryngeal patterns listed in the Classification Manual of Voice Disorders were used as inclusion criteria, only quantitative measures of supraglottic mediolateral compression severity was included in the study design because this pattern has been previously shown to be significantly different between groups with and without pMTD.^9,43 Participants in the control group (mean age: 32.65 years; age range: 20-62 years; 81% women) had no voice issues in the past 6 months, had to score less than 7 out of 40³⁸ on the VHI-10, less than 22 on Part 1 of the VFI, and had to have a laryngeal exam that was functionally, structurally, and neurologically within normal limits.

Data Acquisition Procedures

Ultrasound Cine Videos.

A trained sonographer, board-certified laryngologist with more than 20 years of experience, and voice-specialized speech-language pathologist with 15 years of experience independently confirmed that only target paralaryngeal muscles groups were being captured within the ultrasound video frames at a resolution of 1024x768 pixels. The anterior digastric, mylohyoid, and geniohyoid suprahyoid muscles were identified on ultrasound in trans-midline view by first placing the 4-10 MHz transducer at midline under the chin, starting at the symphysis of the mandible, and then moving the transducer under the chin down towards the neck until the body of the hyoid bone could be visualized. Once the mandible and hyoid bone were identified, the transducer was placed in position so only the trans mid-belly of the anterior digastric, geniohyoid, mylohyoid muscles, as well as the thin platysma muscle (superiorly) were visualized on the screen without capturing the more lateral muscle groups under the chin and base of tongue (Figure 1A). A continuous video loop (cine) was initiated prior to phonation, at which point participants were asked to sustain the vowel /i/ for 3-5 seconds at modal pitch and comfortable loudness, which was also recorded until 1-2 seconds post-vocal offset.

Figure 1: Schematic of Ultrasound Cine Video Acquisition of the Paralaryngeal Muscles.

A) Acquisition of suprahyoid paralaryngeal muscles in the transverse view (anterior digastric, geniohyoid, mylohyoid). B) Acquisition of infrahyoid paralaryngeal muscles in the longitudinal view (thyrohyoid, sternohyoid, sternothyroid). TH=thyrohyoid muscle, SH=sternohyoid muscle, ST=sternothyroid muscle, AD=anterior digastric muscle, GH=geniohyoid muscle. Although the thin belly of the platysma is also present, it takes up less than 1/8 of the cine video frame and thus any significant muscle patterns resulting from platysma movement are negligible. Image partially created in Biorender.com.

The thyrohyoid and sternohyoid infrahyoid muscle groups were then identified by first placing the transducer over the hyoid bone and then moving the transducer inferiorly until only the hyoid bone, thyroid cartilage, longitudinal belly of the thyrohyoid, longitudinal belly of the sternohyoid, and thin layer of the platysma were in view on ultrasound (Figure 1B). The longitudinal sternothyroid muscle belly was also identified using the thyroid cartilage, sternohyoid, and sternum as anatomical landmarks. Cine videos for the infrahyoid muscle groups were recorded in the same manner as for the suprahyoid muscles. Using the hyoid and thyroid as anatomical landmarks at midline also prevented more lateral imaging of the omohyoid, scalene, and sternocleidomastoid muscles.

Each view was captured in triplicate, once before and once after the half hour vocal load challenge. The decision to capture paralaryngeal muscle movement during voice onsets and offsets instead of steady state vowels was based on our previous studies demonstrating similar paralaryngeal muscle stiffness patterns⁹ and acoustic vocal output⁴² between individuals with and without pMTD during steady state vowels. The decision was also supported by the greater laryngeal stiffness levels and laryngeal movement variabilities observed using acoustic proxy methods during voice onsets and offsets in patients with pMTD (e.g., relative fundamental frequency,^44,45 variability of voice onset time⁴⁶).

Laryngoscopy.

Laryngoscopic recordings were acquired on all 42 participants simultaneously with ultrasound videos, at both pre- and post-vocal load time points, in accordance with previously described methods.^9,42,43 Laryngeal movement patterns during six trials of sustained /i/ vowels at comfortable pitch and loudness were obtained with standard flexible nasoendoscope (Olympus distal chip, models ENF-VH, Tokyo, Japan) and xenon light and stored on ImageStream (Portage, Michigan) for subsequent offline supraglottic compression analysis.

Voice Samples.

Acoustic soundwaves of six utterances of vowel /i/ each produced for 3-5 seconds were recorded at a sampling rate of 44.1 kHz in Praat (version 6.1.16) in conjunction with the cine videos and laryngoscopy. Prior to recording, participants were fitted with a MicroMic C250 head mounted condenser microphone at a distance of 4 cm from the lips at 45° angle. The microphone was connected to an M-Audio Air 192, 4 USB C Audio Interface microphone preamplifier and a Dell XPS laptop computer. All voiced segments were captured at modal pitch and loudness.

Vocal Load Task and Self-Perceptual Ratings.

After the pre-vocal load data acquisition with ultrasound, laryngoscopy, and acoustic voice sample, participants rated their vocal effort using a 100 mm visual analog scale. They were also asked to rate their severity of vocal tract discomfort symptoms on the Vocal Tract Discomfort Scale (VTDS).⁴⁷ Once the two self-perceptual ratings were acquired, participants sat 30 cm from a dB SLP meter microphone and read a fiction novel at eighth-grade reading level at 85 dB(A) or greater for half an hour with no breaks. The protocol has previously been applied in the UT Larynx Lab across various studies to induce acute vocal fatigue as safely as possible without causing substantial phonotrauma.^9,42,48 The participants monitored their vocal output with visual cues using DATQ DI-720 USB acquisition hardware and software and were encouraged to increase their vocal intensity when they dipped below the 85 dB level by one of three research coordinators dedicated to the project. Immediately after the vocal load task, participants were asked to rate their vocal effort on the visual analog scale and their vocal tract discomfort symptoms on the VTDS to capture post-vocal load self-perception ratings.

Data Analysis

Optical Flow.

Optical flow analysis of ultrasound cine videos was conducted separately for the suprahyoid (anterior digastric, mylohyoid, and geniohyoid) and infrahyoid (thyrohyoid, sternohyoid, sternothyroid) muscle groups, using the Farneback method via the estimateFlow function in MATLAB 2022b (Mathworks, Natick, Massachusetts, USA) (Figure 2A). Compared to other optical flow methods, the Farneback method is a computationally demanding, dense method of optical flow that incorporates the addition of several techniques that increase suitability for use in noisy environments where motion would normally be difficult to estimate.⁴⁹ To estimate motion vectors, the Farneback algorithm takes the original pair of video frames being analyzed and first estimates the motion of pixels at a lower resolution, then with successively higher and higher resolutions until the original video resolution is met. Lower resolution estimates of motion then inform subsequent higher resolutions of analysis. This analytical method allows the Farneback algorithm to filter out the effects of motion stemming from image grain and static on ultrasound, while tracking larger groups of pixel movement, effectively increasing signal to noise ratio. This also allows for effective pixel tracking over large distances between successive frames in cases where objects move with a substantially high velocity, as is often the case in the paralaryngeal muscles. The estimateFlow Farneback method of optical flow was used at a resolution level of 3, a pyramid scale of 0.50, over 3 iterations, with a pixel neighborhood size of 5, and a pixel filter size of 15.

Figure 2: Overview of Optical Flow Methodology.

A) Optical flow is calculated between consecutive frames of ultrasound cine videos. Vector magnitudes and orientations are calculated for each pixel cluster using the Farneback optical flow method. A pixel cluster’s direction of movement is indicated by the orientation of an arrow on the optical flow quiver plot, and the velocity magnitude is indicated by the length of an arrow’s shaft. Longer shafts represent a farther movement between frames. Optical flow vectors are then filtered based on frame greyscale color values to minimize tracking of off-target, non-paralaryngeal muscle structures. B) Optical flow X (green) and Y (dotted purple) vector components and vector magnitude (black) over the course of a single voice onset and offset. Quiver plots show example optical flow activity during periods of voice onset, steady state, and voice offset. All quiver plots are displayed with a vector scale factor of 5, and a decimation factor of 10 for increased visibility. Shaded error bars indicate standard deviation of vector velocities and components.

Although the Farneback method of optical flow is optimized for tracking fast moving objects in noisy environments, such as in ultrasound cines, images were also filtered based on greyscale color value to minimize tracking of off-target, non-paralaryngeal muscle structures such as cartilage and adipose/epithelial tissue. Videos were color-masked to only include greyscale intensities that maximized isolation of muscular tissue, where black pixels were valued at 0 and the brightest white pixels were valued at 255. Pixel brightness was masked to only include intensities falling between values of 40 and 80, which maximized tracking of the target paralaryngeal muscles. This filtering method ensured that only target suprahyoid and infrahyoid muscle groups captured on the ultrasound videos were being tracked, and that off-target non-muscle structures captured on the ultrasound video, including laryngeal cartilage and adipose/epithelial tissue, were not included in optical flow velocity measures across phonatory time points.

After acquiring motion vector information, optical flow data were aligned to time points of voice onset and offset during each ultrasound recording, when the paralaryngeal muscles are engaged and when they are thought to play the biggest role in vocal hyperfunction and dysfunction patterns^44-46 (Figure 2B). The vector velocity components for each set of muscle groups were used to calculate pixel displacement, including velocity and orientation during the relative time period leading up to voice onset, steady state vowel vocalizations, and relative point to voice offset. To estimate the variability, or muscle movement instability at these time points, the root-mean-square amplitude of muscle pixel displacement was also calculated.

Laryngoscopic Assessment (supraglottic compression severity).

Mediolateral supraglottic compression during phonation, normalized to endolaryngeal width during the steady state vowel production, was used to evaluate group differences at pre- and post-vocal load time points, in accordance with previously validated methods for the pMTD population.⁴³ Main effects of group and vocal load as well as group x vocal load interactions were determined with 2x2 ANOVA and corrected with Bonferroni correction to reduce Type 1 error.

Acoustic Analysis of Voice Samples.

All non-voiced segments were removed from the steady state vowels samples for each participant across the five recorded trials, so that each voice sample had an 18-30 second clip of continuous vocalizations. Batch mode samples were then run on the voice samples using a customized script in Praat (Phonanium CommV) to obtain cepstral peak prominence (CPP) values for each participant for the pre- and post-vocal load time points. Main effects of group and vocal load as well as group x vocal load interactions on CPP measures were determined with 2x2 ANOVA, correcting with Bonferroni correction to reduce Type 1 error.

Self-Perception Ratings of Vocal Effort and Vocal Tract Discomfort.

Each tick mark on the 100 mm vocal effort visual analog scale was measured with a ruler. Sum scores for vocal tract discomfort was calculated using the VTDS. Both sets of numbers were added to Excel spreadsheets and separate 2x2 ANOVAs run on severity of vocal effort and vocal tract discomfort. All ANOVAs were corrected with Bonferroni to account for Type 1 error.

RESULTS

Optical Flow of Paralaryngeal Muscle Movement Velocity and Variability.

Increased optical flow velocities (Figure 3) and variabilities (Figure 4) were observed at both voice onsets and offsets compared to steady state vowels, in both groups, at pre- and post-vocal load time points, separately for trans-midline suprahyoid (anterior digastric, mylohyoid, geniohyoid) and longitudinal left/right infrahyoid (thyrohyoid, sternohyoid, sternothyroid) muscles (see point 0 on the x-axes in Figure 3 and Figure 4). However, there were no group differences on paralaryngeal muscle optical flow velocity or variability patterns during voice onset and offsets, at pre- and post-vocal load time points, or across suprahyoid and infrahyoid muscle groups, between patients with pMTD and vocally healthy controls (see bar graphs in Figure 3 and Figure 4 and p-values in Table 1).

Figure 3: Optical Flow Velocity Pre- and Post-Vocal Load.

A) Pre-vocal load optical flow analysis of paralaryngeal muscle velocities between pMTD and healthy controls during vocal onset and vocal offset in the suprahyoid and infrahyoid muscle groups. Traces show continuous velocities 0.6 seconds before and after voice onset and offset. Bar plots show average velocities over ±0.6 seconds from voice onset and offset. B) Post-vocal load optical flow analysis. No significant differences were found in pixel velocity across transverse midline (suprahyoid) or left and right longitudinal (infrahyoid) views during voice onset or voice offset both pre- and post-vocal load.

Figure 4: Optical Flow Variability Pre- and Post-Vocal Load.

A) Pre-vocal load optical flow analysis of paralaryngeal muscle velocity variability between pMTD and healthy controls during vocal onset and vocal offset in the suprahyoid and infrahyoid muscle groups. Traces show continuous velocity variability 0.6 seconds before and after voice onset and offset. Bar plots show average variability over ±0.6 seconds from voice onset and offset. B) Post-vocal load optical flow analysis. No significant differences were found in pixel velocity variability across transverse midline (suprahyoid) or left and right longitudinal (infrahyoid) views during voice onset or voice offset both pre- and post-vocal load.

Table 1.

p-values for optical flow analysis comparisons of mean velocity and variability at the pre-load and post-load time points between suprahyoid (trans-midline) and infrahyoid (long-right, long-left) paralaryngeal muscles. There were no significant differences found across comparisons.

	Pre-Vocal Load				Post-Vocal Load
	Vocal Onset		Vocal Offset		Vocal Onset		Vocal Offset
	−0.6:0 s	0:+0.6 s	−0.6:0 s	0:+0.6 s	−0.6:0 s	0:+0.6 s	−0.6:0 s	0:+0.6 s
Mean Velocity
Suprahyoid	0.2135	0.27035	0.7227	0.32158	0.29119	0.46252	0.57064	0.45694
Right Infrahyoid	0.60631	0.83786	0.88215	0.81589	0.49109	0.56141	0.36331	0.4549
Left Infrahyoid	0.88664	0.48317	0.88742	0.64481	0.57277	0.97696	0.77528	0.88874
Velocity Variability
Suprahyoid	0.36159	0.15453	0.49857	0.31394	0.69105	0.44904	0.79299	0.85076
Right Infrahyoid	0.63857	0.97136	0.98982	0.73717	0.43149	0.64407	0.33353	0.59933
Left Infrahyoid	0.93417	0.70891	0.86931	0.63557	0.53827	0.93968	0.4982	0.79896

Laryngoscopic Patterns of Mediolateral Supraglottic Compression.

Group analysis showed significant effects of group on mediolateral supraglottic compression (p<0.0001). Although the pMTD group had significantly higher mediolateral supraglottic compression at both pre- (p=0.026) and post-vocal load (p=0.0001) time points, group analysis revealed no effect of vocal load on mediolateral compression (p=0.055). There were also no interaction effects between group and vocal load (p=0.183) (Figure 5).

Figure 5: Mediolateral Supraglottic Compression Severity.

Comparison of mediolateral supraglottic compression severity between pMTD and healthy controls. The pMTD group exhibited significantly more severe mediolateral supraglottic compression than controls at both pre- and post-vocal load time points. (***p<0.0001).

Cepstral Peak Prominence Acoustic Measures.

The pMTD group had significantly lower CPP values at both pre-vocal load (p=0.008) and post-vocal load time points (p=0.001) than the control group. Vocal load did not have a significant effect on CPP (p=0.961), and there were no interaction effects between group and vocal load (p=0.328) (Figure 6).

Figure 6: Cepstral Peak Prominence.

Cepstral Peak Prominence values of control and pMTD groups at pre- and post-vocal load time points. The pMTD group exhibited significantly lower CPP values at pre-vocal load (*p<0.05) and post-vocal load time points (***p<0.0001) compared to the Control group.

Self-Perceptual Measures of Vocal Effort and Vocal Tract Discomfort Severity.

Participants in the pMTD rated their vocal effort (p<0.0001) and vocal tract discomfort severity (p<0.0001) significantly higher than participants in the control group. Both groups felt adequately vocally loaded with the half hour loud reading challenge, which was supported by significantly higher ratings of vocal effort (p=0.0002) and vocal tract discomfort severity (p=0.0007) at the post-vocal load time point. No interaction effects were noted between group and vocal load time point for either vocal effort (p=0.557) or vocal tract discomfort severity (p=0.926) (Figure 7).

Figure 7: Vocal Effort and Vocal Tract Discomfort.

Comparison of self-perceptual ratings of vocal effort and vocal tract discomfort. The pMTD group exhibited significantly higher vocal effort and vocal tract discomfort at pre- and post-vocal load time points (***p<0.0001). Both groups reported significantly more vocal effort and vocal tract discomfort at post-vocal load compared to pre-vocal load (*p<0.05, **p<0.01).

DISCUSSION

The goals of this study were to implement optical flow methods to ultrasound videos of the paralaryngeal muscles to capture phonatory paralaryngeal movement velocity and variability patterns and compare these movement patterns between groups with and without a diagnosis of pMTD. Optical flow measures captured greater paralaryngeal muscle movement with voice onsets and offsets, compared to pre-phonation time points, steady state vowels, and post-phonation, in both suprahyoid and infrahyoid muscle groups, as would be expected with vocal productions. These findings demonstrate the utility of using optical flow to characterize paralaryngeal muscle biomechanics with phonation. However, the level of phonatory movement velocity and variability were similar between groups, suggesting similar muscle movement patterns in the paralaryngeal muscles occur with phonation in individuals with and without a diagnosis of pMTD, at least with modal pitch and loudness levels. In contrast, patients with pMTD had significantly more mediolateral supraglottic compression and more vocal perturbations (as measured via CPP) with sustained vowels, and reported significantly higher vocal effort and vocal tract discomfort severity. These findings suggest the intrinsic laryngeal system may play a larger role in vocal dysfunction patterns in the context of pMTD than the paralaryngeal muscles, at least at modal pitch and loudness.

Similarities in phonatory optical flow patterns of the paralaryngeal muscles between groups parallel the inconsistent findings found across previous studies^16,50 using methods like paralaryngeal palpation⁵¹ and electromyography to quantify paralaryngeal muscle “hyperfunction.”^19-21,50,52 Common interpretation of these findings has been that these methods lack sensitivity and precision to capture paralaryngeal hyperfunction. However, the inconsistency of these approaches, in the context of lack of group differences on phonatory paralaryngeal movement patterns using objective, quantifiable metrics like optical flow found in this current study, suggests the issue may have more to do with the erroneous attribution of paralaryngeal hyperfunction to pMTD than the imprecision of previous methods to define paralaryngeal muscle hyperfunction.

The lack of group differences on phonatory movement velocity and variability of the paralaryngeal muscles using optical flow also makes sense from a physiological perspective. The primary role of the paralaryngeal muscles that support the larynx is aerodigestive in nature⁵³ and when these muscle groups are involved in voicing, they typically result in pitch changes.^8,54,55 However, pitch deviations are typically not a clinical presentation central to a pMTD diagnosis. In fact, a common feature in patients with pMTD is that their voices sound less aberrant compared to other voice disorders^40,42 and many patients with pMTD have more of an issue with the feel^41,42 of their voice than the sound of their voice. These clinical presentations were supported by the significantly higher vocal effort and discomfort ratings patients in the pMTD group experienced and the CPP values that fell within the normal range⁵⁶ in this cohort.

The suprahyoid muscle group works together to raise the larynx while the infrahyoid muscle groups lower the hyolaryngeal complex. However, it is possible that certain muscles within the suprahyoid and infrahyoid muscle groups play a larger role than others, and by lumping these muscles by their role in raising and lowering the larynx for optical flow analysis, the role of single muscles on pMTD clinical presentation may have been missed. However, it is highly unlikely since these muscle groups rarely work in isolation. Another possible reason for the lack of findings in this present study could be that steady state vowels at modal pitch and loudness used to acquire optical flow measures with vocal onsets and offsets may not be physiologically demanding enough to elicit different phonatory paralaryngeal muscle patterns in the pMTD cohort, compared to healthy voice users. Future work investigating biomechanical patterns in separate paralaryngeal muscles, especially with increased vocal projection, and at different vocal pitches, is needed to confirm or refute whether there are differences in paralaryngeal muscle patterns between individuals with and without pMTD, especially in the context of vocal intensity and pitch changes. Finally, although supraglottic compression was assessed in this study, vocal fold configurations were not analyzed. Previous work has suggested an effect of the paralaryngeal muscles on vocal fold configurations,⁷ and as such, relationships between vocal fold configuration and paralaryngeal muscle patterns also require elucidation in future investigations with optical flow methods at both modal pitch and pitch changes.

CONCLUSION

Optical flow is a quantifiable, objective method that can be applied to ultrasound videos for the study of paralaryngeal muscle velocity and variability movement patterns in the suprahyoid and infrahyoid muscle groups with phonation. In this study, standard clinical metrics (acoustic, laryngoscopic, self-perceptual) were able to distinguish patients with pMTD form vocally healthy controls. However, there were no movement pattern differences with vocal onsets and offsets between subjects with and without pMTD using optical flow analysis for either suprahyoid or infrahyoid muscles. The lack of group differences using objective and quantifiable methods like optical flow to study phonatory paralaryngeal movement patterns in the suprahyoid and infrahyoid muscle groups challenges clinical wisdom that vocal hyperfunction in the paralaryngeal muscles occurs and contributes to vocal dysfunction in patients with pMTD,^{1-3,10,11,14,15} at least at modal pitch and loudness.

Data Availability:

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.