Effect of Vocal Fold Implant Placement on Depth of Vibration and Vocal Output

Abstract

Objective:

Most type 1 thyroplasty implants and some common injectable materials are mechanically stiff. Placing them close to the supple vocal fold mucosa can potentially dampen vibration and adversely impact phonation, yet this effect has not been systematically investigated. This study aims to examine the effect of implant depth on vocal fold vibration and vocal output.

Study Design:

Computational simulation.

Methods:

Voice production was simulated with a fiber-gel finite element computational model that incorporates a three-layer vocal fold composition (superficial lamina propria, vocal ligament, thyroarytenoid muscle). Implants of various depths were simulated, with a “deeper” or more medial implant positioned closer to the vocal fold mucosa and replacing more muscle elements. Trajectories of surface and within-tissue nodal points during vibration were produced. Outcome measures were the trajectory radii, fundamental frequency (F₀), sound pressure level (SPL), and smoothed cepstral peak prominence (CPPS) as a function of implant depth.

Results:

Amplitude of vibration at the vocal fold medial surface was reduced by an implant depth of as little as 14% of the total transverse vocal fold depth. Increase in F₀ and decrease in CPPS were noted beyond 30% to 40% implant depth, and SPL decreased beyond 40% to 60% implant depth.

Conclusions:

Commonly used implants can dampen vibration “from a distance,” ie, even without being immediately adjacent to vocal fold mucosa. Since implants are typically placed at depths examined in this study, stiff implants likely have a negative vocal impact in a subset of patients. Softer materials may be preferable, especially in bilateral medialization procedures.

INTRODUCTION

The treatment of unilateral vocal fold paralysis with injection laryngoplasty or type 1 thyroplasty places foreign material into or adjacent to the paralyzed vocal fold. While most of the commonly used injectable and thyroplasty implant materials have acceptable biocompatibility from a safety standpoint, the impact of implant placement on the vibratory properties of the vocal fold mucosa is not well known. For instance, calcium hydroxyapatite paste, one of the most commonly used injectable materials, is quite stiff.¹ Virtually all materials used as thyroplasty implants are also stiff. Placing these materials next to the supple vocal fold mucosa can potentially dampen vocal fold vibration, a phenomenon that is anecdotally observed clinically but not extensively studied. Dampening of vocal fold vibration by implants is strongly suggested by the work of Zhang et al.² They showed in an excised human larynx phonation model that a soft silicone implant produced more favorable acoustic outcomes than a silastic implant. The implication was that the stiffer silastic implant impeded vocal fold vibration, but the mechanism was unclear. Since silastic implants are widely used in type 1 thyroplasty, further investigation of the extent to which such implants impede vocal fold vibration and how they do so is warranted.

A key concept in understanding how implants affect vocal fold vibration is the depth of vibration. In the absence of an implant, the flow-induced self-sustained oscillation of the vocal fold tissue propagates from the surface into the tissue for some depth. This depth of vibration varies by subglottal pressure and by tissue property. Loud phonation (high subglottal pressure) involves a greater depth of vibration than soft phonation (low subglottal pressure), and stiffer vocal fold tissue has a lesser depth of vibration than softer tissue. Estimation of the depth of vibration is complicated by the fact that the vocal fold tissue is not homogeneous. The vibration mainly involves the superficial lamina propria, but the vocal ligament and the thyroarytenoid muscle also engage in vibration depending on the vocal register and vocal intensity.³ A recent study using computational simulation showed that the effective depth of vibration ranged between 15% to 55% of the total anatomical depth of the vocal fold and was affected by the layered structure of the vocal fold.⁴ Since implants are typically placed adjacent to the muscle (in thyroplasty) or into the muscle (in injections), implants may negatively impact vocal fold vibration and the voice produced if the depth of vibration reaches the muscle. However, the degree to which implants exert this effect is unknown.

The goal of this study was to determine the effect of implant placement on the depth of vibration and on voice quality. The depth of vibration was quantified and visualized as the trajectory of points within the tissue as simulated using a fiber-gel finite element computational model of the vocal folds. The following hypotheses were tested: 1) There is a finite depth at which a stiff implant will negatively impact phonation, and 2) muscle compression by the implant negatively impacts phonation.

METHODS

Computational Model

The fiber-gel finite element model of the voice simulation software VoxInSilico was used as the investigative tool.^4–8 The acoustic output and behavior of the finite element model have been validated in previous studies.^7,8 An earlier version of the model was used to simulate the effect of varying depths of cordectomy.⁹ The same computational model was used to investigate the effective depth of vocal fold vibration in a previous work,⁴ where the intra-tissue nodal trajectories were in reasonable agreement with those observed in excised human hemilarynges. All simulations were performed with symmetric vocal fold models. The finite element model of the right vocal fold is shown in Figure 1. A 7 (transverse) × 6 (superior-inferior) element mesh was used in the coronal plane, and the anterior-posterior length was discretized into 15 elements (7 × 6 × 15 grid). To investigate the effect of mesh density on acoustic output, the simulations were repeated with a mesh with twice the grid resolution (14 × 12 × 30 grid) and a mesh with thrice the grid resolution (21 × 18 × 45 grid).

Fig. 1.

Schematic of finite element vocal fold model used in VoxInSilico. A right vocal fold is shown.

The coronal cross-section consisted of three tissue layers: superficial layer of the lamina propria (SLLP), vocal ligament, and muscle, which were extended along the entire vocal fold length. Compared to the earlier model,⁹ the muscle layer was expanded medially to help model the presence of a stiffer conus elasticus extending inferiorly from the ligament. The length of the vocal folds was 1.6 cm. The vertical thickness on the medial and lateral surfaces was 0.7 and 1.5 cm, respectively. The transverse depth varied along the length of the vocal fold and followed the curvature of the thyroid cartilage boundary. The depth was 1.53 cm at the posterior end and 0.21 cm at the anterior end.

Prephonatory vocal fold posturing was accomplished by control of activation levels for the intrinsic laryngeal muscles, including the cricothyroid (CT), lateral cricoarytenoid (LCA), thyroarytenoid (TA), and interarytenoid (IA) muscles.⁵ Two muscle activation conditions were tested for this study: CT – 30%, TA – 30% (“3030 condition, low pitch”); and CT – 60%, TA – 40% (“6040 condition, higher pitch”). Activation levels for LCA and IA were held constant at 26% and 30%, respectively, for both conditions. These are typical for achieving normal speech-type phonation,¹⁰ and further justification is provided in a prior study.⁴ The viscoelastic properties of the three layers were governed in part by these muscle activations, as detailed in Palaparthi et al.⁴ For the 3030 condition, the corresponding longitudinal shear moduli were 5.58 kPa for the SLLP, 7.18 kPa for ligament, and 37.3 kPa for muscle. For the 6040 condition, the longitudinal shear moduli were 20.3 kPa for the SLLP, 42.1 kPa for ligament, and 64.2 kPa for muscle. The transverse shear moduli for all three layers were set at 0.5 kPa for both conditions. The voice simulation included the effect of a vocal tract from the lungs to the lips. Vocal fold vibration was simulated for 0.4 sec at three subglottal pressures: 1.0, 1.5, and 2.0 kPa, for each of the models below. The prephonatory glottal width was set to zero at the vocal processes for all simulations. This was to eliminate glottal gap size as a variable, since the goal of the study was to investigate the effect of implant depth independent of glottal gap size.

Vocal Fold Implant Models

The models tested two independent variables: presence of a stiff implant and effect of muscle compression. The following models were employed:

• Model 1: Vocal fold without implant (Fig. 2A)

• Model 2: Vocal fold with stiff implant (Fig. 2B)

• Model 3: Model 2 with correction for muscle compression (Fig. 2C)

Fig. 2.

Models employed for the current study. (A) Normal vocal fold. (B) Vocal fold with progressive depths of a stiff implant, in gray. (C) Model in B, with correction for muscle compression. Only the 29% depth implant is shown for clarity.

Model 1 served as the control. Model 2 tested the effect of implant alone. The implant was modeled by fixing the displacement of the desired nodes to simulate a stiff implant in that portion of the vocal fold. Implant depth (transverse size) was varied by progressively fixing more columns of nodes medially, resulting in implant depths of 14%, 29%, 43%, 57%, and 71% as a percentage of the total vocal fold transverse depth. Model 3 combined the effects of implant and muscle compression. Muscle compression was modeled by scaling muscle material properties (Young’s modulus, transverse shear modulus, longitudinal shear modulus) by a factor of 2 in the muscle elements immediately surrounding the implant elements. The scaling factor of 2 was deemed a reasonable order of magnitude to account for the strain imposed on muscle by the implant, as no literature data regarding muscle deformation by an implant is available for reference. The baseline Young’s modulus was 2 kPa for isotropic incompressible gel, based on shear modulus of 0.5 kPa.

Outcome Variables

Vocal fold model vibratory motion was tracked and compared between cases. The positions of all nodes in the model were recorded during steady state vibration and were plotted as nodal trajectories.¹¹ The trajectories were then analyzed using a custom MATLAB script to determine the effective trajectory radius for each of the seven nodes along the medial surface (from superior to inferior, Fig. 2A) in the mid-membranous plane (coronal section number 8 out of the 15 anterior-posterior sections). The algorithm finds the area enclosed by each trajectory boundary then equates that area to a circular area (πr²) and solves for the radius r.

A comparison of basic acoustic and voice quality parameters was also made. Fundamental frequency (F₀) and sound pressure level (SPL) were calculated from output pressure (sound) waveform in VoxInSilico. Smoothed cepstral peak prominence (CPPS) is associated with the degree of regularity in the acoustic signal and has been shown to be a robust measure of dysphonia.^12–16 The CPPS was calculated from the glottal flow waveform using built-in functions in Praat.^17,18

RESULTS

Nodal Trajectories

A schematic of trajectories for all nodes in the vocal fold models as a function of implant depth, at a subglottal pressure of 1.0 kPa, is shown in Figure 3 for the low pitch condition. The trajectories were greater at the super-omedial vocal fold edge than laterally or inferiorly, consistent with surface measurements in excised larynges.^19–21 The trajectories also decreased with greater transverse depth. As the implant was placed more medially, amplitude of vibration on the medial surface was visibly reduced, as evidenced by the smaller trajectory areas. By 71% medialization (bottom right in trajectory plots), vibration completely damped out. Trajectories for the muscle compression cases are not plotted, as they were very similar.

Fig. 3.

Vibration trajectories at all nodes in the mid-vocal fold coronal section (section number 8 out of the 15 anterior-posterior sections), with simulation done at low pitch and subglottal pressure of 1.0 kPa.

A plot of trajectory radius along the medial surface nodes for all cases confirmed that trajectory steadily decreased as implant medialization increased for all points along the medial surface (Fig. 4). Muscle compression produced slightly more damping of vibration for each case, particularly at the three inferior nodes (15.9% on average at node 7). Trajectories and trajectory radii for the higher pitch condition and higher subglottal pressures were similar.

Fig. 4.

Vibration trajectory radii of the seven nodes along the vocal fold medial surface. Shown for various implant depths, expressed as percentages of the total transverse vocal fold depth. Data shown for the low pitch condition, with subglottal pressure of 1 kPa.

Acoustic Parameters

Figure 5A/B show fundamental frequency (F₀) as a function of the depth of implant placement for both muscle activation conditions with and without muscle compression (Fig. 5A) and at the two higher subglottal pressures (Fig. 5B). The F₀ without implant present was approximately 100 Hz for the low pitch condition and 130 Hz for the higher pitch condition, reflecting the male vocal fold dimensions used in the model. F₀ generally increased as a function of implant depth, although the increase was minimal for implant depths below 30% for the low pitch condition. The increase in F₀ was more obvious for the high pitch condition. Muscle compression had a minimal influence on F₀.

Fig. 5.

(A,B) Fundamental frequency F₀, (C,D) sound pressure level (SPL), and (E,F) smoothed cepstral peak prominence (CPPS) as a function of the depth of implant placement. Panels A, C, E show results from simulation at 1 kPa subglottal pressure. Panels B, D, F show results from simulations at 1.5 and 2.0 kPa subglottal pressures. Higher CPPS values correspond with better acoustic quality.

Figure 5C/D show sound pressure level (SPL) as a function of the depth of implant placement. Generally, SPL began to decrease above 30% implant depth, with a more notable decrease for the higher pitch condition. The uptick at 57% implant depth in the low pitch condition is anomalous (see Discussion). The SPL stayed within about a 5 dB range for all conditions shown in Figure 5C/D up to about 40% implant depth. Including muscle compression yielded similar results.

Smoothed cepstral peak prominence (CPPS) values for all cases are shown in Figure 5E/F. For the low pitch condition with a subglottal pressure of 1.0 kPa, CPPS decreased above 30% implant depth (Fig. 5E). With higher subglottal pressures, the CPPS decreased as a function of implant depth for the low pitch condition (Fig. 5F). For the higher pitch condition, CPPS decreased above 30% implant depth but then increased at 57% depth or above, reaching near normal values at 71% implant depth. Muscle compression again yielded only minor differences in CPPS.

The effect of modeling the vocal fold with higher resolution finite element mesh on acoustic output is shown in Figure 6. While the absolute F₀ showed mesh dependency, the change in F₀ as a function of implant depth is similar, with greater implant depths leading to higher F₀. SPL and CPPS, which changed little at implant depths at or below 30% to 40% in simulations performed with the default mesh, appeared more sensitive to implant depth and decreased with implant depth as little as 14%.

Fig. 6.

F₀, SPL, and CPPS as functions of implant depth, using different grid sizes for the finite element models.

DISCUSSION

This computational study showed that placement of a stiff medialization implant in the paraglottic space dampens vocal fold tissue vibration and can negatively impact vocal output. The amplitude of vibration at the vocal fold medial surface was reduced by a lateral implant depth of as little as 14% of the total vocal fold depth in the transverse dimension. However, the functional impact was not apparent till the implant depth increased beyond 30% of the total depth. At 40% and greater implant depth, there was a notable increase in F₀ and decrease in acoustic quality as reflected in the CPPS. At depth greater than 60%, the SPL began to drop. We should note that the change in SPL and CPPS as a function of implant depth was somewhat dependent on the resolution of the finite element mesh used, so it is possible that SPL and CPPS may degrade at implant depths less than 40% to 60%. As a point of reference, the depth of thyroplasty implants has been measured to range from 40% to 70% of the total depth in live patients,²² and the depth of injected calcium hydroxyapatite boluses ranged from 70% to 80% in a study using cadaveric human larynges.²³ It is therefore likely that the use of a stiff implant has an unfavorable effect on vocal output in at least a subset of patients.

The mechanism by which F₀ increases with greater implant depth can be explained on the basis of the depth of vibration. To a first approximation, F₀ is inversely related to the depth of vibration.²⁴ Fundamental frequency is proportional to $\sqrt{\frac{k}{m}}$, where k is material stiffness and m is mass. The increase in F₀ as implant medialization increased was likely due to two factors: 1) the decrease in mass of the portion of the fold allowed to vibrate; and 2) the increase in effective stiffness as the vibrating portion decreased in medial-lateral length.

We wish to emphasize that the trend of acoustic parameters as a function of implant depth is far more meaningful than the absolute numerical values of those parameters from simulations using a relatively simplified model, since those values depend heavily on material properties and other assumptions in the model. One notable exception to the trend was the uptick in SPL at 57% implant depth (Fig. 5C/D). A potential explanation is that this occurred at a point of resonance where standing waves existed in the tissue with that particular implant. The increased CPPS at various points (Fig. 5E/F) can potentially be explained as mode entrainment that increased periodicity under certain conditions. Because of the nonlinear nature of the model, these phenomena are often unpredictable.

The current results are consistent with those of a prior study of thyroplasty in excised human larynges. Zhang et al.² showed that the vocal output was very sensitive to the insertion depth of a stiff implant (Silastic), including a significant increase in F₀ with full insertion. In contrast, softer grade silicone implants were more forgiving in that the acoustic quality was less sensitive to implant insertion depth. The current study extends the prior work with data on tissue movement within the vocal fold and provides a mechanistic explanation of the experimental data. This work also used more finely graded levels of implant depth, which led to an important finding: There is an implant depth threshold below which normal voice can be achieved. If the depth of insertion is less than about 30% of the total transverse depth, even a stiff implant should not negatively impact vocal output.

This study directly demonstrated the mechanism for vocal fold mucosa “stiffness” following medialization procedures. Stiffness caused by thyroplasty has been inferred in excised larynx²⁵ and in vivo canine²⁶ experiments. It is also seen in some patients following type 1 thyroplasty or injection laryngoplasty and manifests as reduced vibrational amplitude on stroboscopy. This work shows that the reduction in amplitude can be explained on the basis of a biomechanical effect from the implant, rather than actual stiffening of the vibratory tissue.

Muscle compression by the implant was not found to have a significant effect on vibration or vocal output. Testing the effect of muscle compression was motivated by the observation that calcium hydroxyapatite injection was associated with significant vocal fold muscle compression in histology sections of cadaveric larynges.²³ The absence of a large effect from muscle compression per se indicates that the chief determinant of negative vocal impact from implant is the depth of insertion, and not compression of the soft tissue medial to the implant.

We should note that the glottal width was kept constant at a minimum in all cases simulated in this study, so implant depth was varied while good glottal closure was maintained. In other words, the intent of the study was to probe the role of implant depth independent of glottal closure. This would be difficult to achieve with real larynges and is facilitated by computational simulation. Prior experiments using excised larynx² or in vivo phonation²⁶ have demonstrated beneficial effects of progressive implant insertion, but those benefits largely relate to improved glottal closure. Our simulations were performed with optimal glottal closure at baseline, so the benefits seen in the prior experiments were not expected in this study.

This study has several implications. First, it underscores the importance of real-time auditory feedback with implant adjustment during thyroplasty. Intraoperative laryngoscopy provides visual guidance on the position of the vocal fold edge but does not convey the vocal quality. A vocal fold fully medialized to the midline may require a larger-than-optimal depth of implant insertion, which can cause increased F₀ and reduced SPL, the combination of which conveys a quality of strain. So a slight visual under-correction may be preferable in some cases. Second, when a stiff material such as calcium hydroxyapatite is used for injection laryngoplasty, it should be targeted as laterally as possible to minimize potential damping of the vibration.

The third and perhaps most important implication of this study has to do with bilateral thyroplasties for age-related vocal fold atrophy. As the vocal fold muscles atrophy, the tissue envelope medialized by a thyroplasty implant (“musculomucosal envelope”) becomes thinner. It is likely that the percentage of total depth taken up by the implant will be proportionally higher, with more negative impact on vibration and vocal output. The prediction is that, if stiff implants are placed bilaterally, it may be difficult to achieve an optimal vocal outcome. For the treatment of bilateral vocal fold atrophy, softer materials may be preferable over silastic implants.

This study has two main caveats. First, as with any computational modeling, the results of simulation are only as good as the model itself. Various aspects of the model have been validated.^7,8 In addition, dynamics of vibration at the vocal fold medial surface calculated using the same approach as this study have been compared to those measured in an excised human hemilarynx model and showed good qualitative agreement.⁴ These give us confidence that the model output is a reasonable estimate of reality. Second, our simulations were performed with symmetric vocal fold morphologies, ie, with bilateral implants. Some of the simulated effects may over-estimate those from unilateral implants, which is the far more common clinical scenario. Nevertheless, we believe the insights from this work are still applicable on a qualitative basis.

CONCLUSION

Commonly used stiff thyroplasty implants and injectable materials to medialize vocal folds may negatively impact vocal output in a subset of patients. Patients most likely to be impacted are those with atrophied vocal folds, for example in chronic denervation or severe age-related atrophy. Softer materials may be preferable, especially in bilateral medialization procedures.