Numerical and Experimental Investigations on Vocal Fold Approximation in Healthy and Simulated Unilateral Vocal Fold Paralysis

Round 1

Reviewer 1 Report

Review paper:

Numerical and experimental investigations on vocal fold approximation in healthy and simulated unilateral vocal fold paralysis

Content is a method development on how to develop an EM model from MRI data. The natural frequencies are determined. The developed model allows studies depending on the material parameters for voice formation.

The MRI data are based on studies of rabbit larynges. COMSOL is used as the FEM code. The method is described in detail.

Some remarks:

- Figure 4 shows the computational grid. Has a grid study been done to show the independence of the solution from the grid size?

- An important point are the material parameters for the biological material. Literature references are given here. More information is needed to understand the publication. It is an essential part of the simulation.

Finally, there are a few comments to be made about the medical usability of the research methodology. The methodology is very useful and helpful for voice research. However, vocal fold vibration must always be seen in interaction with flow. This is where the acoustic source terms for the voice are generated. An essential issue is also the contact problem in vocal fold vibration.  Here, slightly different vocal folds can synchronize in frequency again. Another issue is the mucosal wave.

It would also be important to analyze the examinations in connection with the acoustics. The visual images have only limited significance here.

But that is still a further research work.

I am for a publication with minor changes to the above comments.

Author Response

Reviewer 1

Response: Thank you for the great and helpful suggestions. We have revised the manuscript accordingly.

Comment 1: Figure 4 shows the computational grid. Has a grid study been done to show the independence of the solution from the grid size?

Response: Details for mesh-independent study have been added to section 2.2.1.

Comment 2: An important point are the material parameters for the biological material. Literature references are given here. More information is needed to understand the publication. It is an essential part of the simulation.

Response: We have addressed this in several places throughout the manuscript. Firstly, in the introduction we broadly define the structure of the vocal fold. Secondly, in section 2.2.1 we discuss the vocal fold constitutive model in terms of stress and strain. Finally, in section 3.2 we pointed out that Young’s moduli we used are in the range of the previous experimental data.

Comment 3: Finally, the methodology is very useful and helpful for voice research. However, vocal fold vibration must always be seen in interaction with flow. This is where the acoustic source terms for the voice are generated. An essential issue is also the contact problem in vocal fold vibration.  Here, slightly different vocal folds can synchronize in frequency again. Another issue is the mucosal wave.

It would also be important to analyze the examinations in connection with the acoustics. The visual images have only limited significance here.

Response: We agree with the reviewer that fluid-structure interaction (FSI) is very important in vocal fold vibration. Our group has the capability to perform high-fidelity FSI simulations, which were represented by our existing publications. However, the present study is focused on the integrated experimental/computational study of the UVFP, where we aim to first explore important issues such as different vocal fold configurations, fundamental frequency and mode, and experimental validation of the computational model.  FSI will be pursued in the future study. On the other hand, it has been demonstrated that vocal fold vibration is highly related to its eigenfrequency (Berry, 2001; Yin & Zhang, 2013). Once our integrated study is established, our future work can address other critical aspects such as FSI, contact, mucosal wave, and acoustic parameters. We have clarified this point in the limitations in the revised manuscript.

Clinical applications for this model have now been addressed in section 4. Briefly, we plan to use this subject specific approach of modeling to optimize surgical planning for type I thyroplasty procedures.

Author Response File: Author Response.pdf

Reviewer 2 Report

The Authors present a study about the capability of a FEM model of rabbit larynges to predict the frequency of healty vocal folds and for the case of a uni-lateral paralysis. Therefore, they first performed in-vivo phonation tests with rabbits whose vocal folds were adapted by a Cricothyroid suture technique to mimick normally and one-sided adducted positions of the vocal folds. During the experiments, the oscillation frequencies were measured using high-speed recording with a pediatric endoscope. Simultanously, three excised rabbit larynges were scanned by MRI and volumetric FEM models were prepared. A subsequent model analysis yielded the eigenmode and eigenfrequency of the vocal folds which the authors attributed to the vibration of the vocal folds.

The manuscript is adequately structured and clearly written. The results show qualitatively good agreement between experimental and numerical models. However, there are some point that have to be addressed before it can be considered for publication.

• The authors mentioned, that the larynges from the rabbits used in the in-vivo measurements could not be segmented due to inflammation and damages caused by the phonation tests. So why did they not use only ex-vivo larynges for both, the experiments and modal analysis? What advantages of performing in-vivo measurements without stimulation existed while accepting the significant shortcoming to compare the experimental results with numerical ones from different larynges? Please clarify.
• Please provide more information about the FEM model (number of finite elements in total and for the different parts, shape of volume elements), the material model (extension of table 2 with data of all parts) and the modal analysis (method, numerical parameters).
• How were the values of the tissue model in table 2 found or defined? From literature or from tensile tests?
• Did the authors checked the results to be independent for grid resolution?
• The authors only show the eigenmode with predominantly medial-lateral motion. They should however provide an overview of all relevant eigenmodes with the associated frequencies. Additional, supplement movies of the eigenmode dynamics would improve the outcome of this study.
• How large was the oscillation amplitude of the shown eigenmode compared with the measured vibration amplitudes?

Author Response

Reviewer 2

Response: Thank you for the great and helpful suggestions. We have revised the manuscript accordingly.

Comment 1: The authors mentioned that the larynges from the rabbits used in the in-vivo measurements could not be segmented due to inflammation and damages caused by the phonation tests. So why did they not use only ex-vivo larynges for both the experiments and modal analysis? What advantages of performing in-vivo measurements without stimulation existed while accepting the significant shortcoming to compare the experimental results with numerical ones from different larynges? Please clarify.

Response: The advantage for obtaining HSV in vivo is that we are able to capture vibratory data from tissues that are responding under normal physiological conditions. If HSV is performed ex vivo, extremely bright light is required to capture at the necessary frame rate, and this contributes to tissue degradation. To avoid this, tissues can be preserved, but in either case vibratory properties will be altered. Due to the duration of the MRI scans to obtain appropriate resolution for computational modeling (approximately 4 hours per scan) we are not able to gather this data in vivo due to concerns of protracted anesthesia, movement during scans, and general welfare concerns. For these reasons we believe that use of in vivo HSV and ex vivo MRI scanning is the most viable and accurate approach. We’ve revised the language in sections 2.2 and 4 to reflect this.

Once we have a more complete computational model we will use lower resolution in vivo MRI scans to complement in vivo HSV.

Comment 2: Please provide more information about the FEM model (number of finite elements in total and for the different parts, shape of volume elements), the material model (extension of table 2 with data of all parts) and the modal analysis (method, numerical parameters).

Response: We have included these details accordingly in section 2.2.1, and Table 2.

Comment 3: How were the values of the tissue model in table 2 found or defined? From literature or from tensile tests?

Response: Firstly, based on the previous experimental studies (Rousseau et al., 2004; Thibeault, Gray, Bless, Chan, & Ford, 2002), the Young’s modulus of the vocal fold cover ranges from 0.2kPa to 20kPa. This property in our current work falls into this range. Secondly, with the tissue property settings in Table 2, we can achieve a dominant eigenfrequency around 600Hz for the HP condition. It is consistent with the average frequency across the rabbit larynx samples in two previous studies from our group (Chang et al., 2016; Novaleski et al., 2016).  We have added this to the discussion.

Comment 4: Did the authors checked the results to be independent for grid resolution?

Response: Details for the mesh-independence study have been added to section 2.2.1.

Comment 5: The authors only show the eigenmode with predominantly medial-lateral motion. They should however provide an overview of all relevant eigenmodes with the associated frequencies. Additionally, supplement movies of the eigenmode dynamics would improve the outcome of this study.

Response: An overview of the relevant eigenmodes and associated frequencies have been included as supplementary material. Those modes are inconsistent with the vibration modes observed in the experiment and thus are not discussed further.

Comment 6: How large was the oscillation amplitude of the shown eigenmode compared with the measured vibration amplitudes?

Response: The amplitude in eigenmode is an arbitrary number, so unlike an FSI simulation, the absolute value of eigenmode amplitude cannot be compared with the measured vibration. The relative value of the displacement in the eigenmode represents the vibration pattern that is discussed in the study.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Supplement material:

The Supplement material is relevant and increases the quality of the manuscript. However, the authors should include a larger coordinate system in pictures. Furthermore, the vocal fold that presents the paralyzed once should be indicated in the caption of the corresponding video.

Author Response

Response: Thank you for the great and helpful suggestions. We have revised the supplemental material accordingly.

Comment 1: The Supplement material is relevant and increases the quality of the manuscript. However, the authors should include a larger coordinate system in pictures. Furthermore, the vocal fold that presents the paralyzed once should be indicated in the caption of the corresponding video

Response: We have revised the supplemental material based on the reviewer’s comment. Larger coordinate system in pictures and caption to present the paralyzed side have been included.