Impact of Respiratory Gating on Hemodynamic Parameters from 4D Flow MRI

Round 1

Reviewer 1 Report

The authors have performed an interesting study investigating the impact of respiratory motion on the in vivo computation of hemodynamic parameters using 4D flow MRI data from 15 healthy volunteers. Statistical analyses were performed to quantitatively assess the differences between hemodynamic parameters computed from 4D flow MRI data acquired with and without respiratory gating, and the potential impact of geometry segmentation on the statistical differences between hemodynamic parameters was also investigated. The authors present their hypothesis and discuss their findings in a clear and exhaustive way.

 I have a few comments listed below:

1. My main issue is the lack of information and details regarding the computation of hemodynamic parameters from the acquired 4D flow MRI velocity data. The authors provide a reference for the segmentation and FE mesh generation, but nothing is said, nor references are explicitly provided, regarding the hemodynamic characterization. For example, how were velocity data computed in the FE mesh nodes? How was WSS estimated from in vivo data? Please revise the corresponding section in Materials and Methods adding the missing information or references to previous studies.
2. A characteristic length of 1mm was chosen for the FE mesh used to compute the hemodynamic parameters. Is this choice based on specific considerations? Have the authors performed some “sensitivity” study to ensure that such characteristic length allows for an accurate computation of the spatial derivatives of velocity and, in turn, of WSS, vorticity, ecc..?
3. In the Introduction, the authors mention helicity among the hemodynamic parameters widely studied in literature. A method for quantifying helical flow in vivo was proposed in Morbiducci et al., 2009, precisely employing 4D flow MRI data of human thoracic aorta. Given the increasing interest of the scientific community in studying helical flow and its beneficial role in human arteries, have the authors considered to include this hemodynamic parameter in their studies?

Author Response

The authors have performed an interesting study investigating the impact of respiratory motion on the in vivo computation of hemodynamic parameters using 4D flow MRI data from 15 healthy volunteers. Statistical analyses were performed to quantitatively assess the differences between hemodynamic parameters computed from 4D flow MRI data acquired with and without respiratory gating, and the potential impact of geometry segmentation on the statistical differences between hemodynamic parameters was also investigated. The authors present their hypothesis and discuss their findings in a clear and exhaustive way.

 

Thank you for your comments.

 

I have a few comments listed below:

 

1. My main issue is the lack of information and details regarding the computation of hemodynamic parameters from the acquired 4D flow MRI velocity data. The authors provide a reference for the segmentation and FE mesh generation, but nothing is said, nor references are explicitly provided, regarding the hemodynamic characterization. For example, how were velocity data computed in the FE mesh nodes? How was WSS estimated from in vivo data? Please revise the corresponding section in Materials and Methods adding the missing information or references to previous studies.

 

Regarding the first point, we add references to the computation of hemodynamic parameters previously developed using a least-square projection method based on finite elements. The velocity values were transferred to each node of the mesh from the 3D PC-MRI data sets using cubic interpolation.

 

 

2. A characteristic length of 1mm was chosen for the FE mesh used to compute the hemodynamic parameters. Is this choice based on specific considerations? Have the authors performed some “sensitivity” study to ensure that such characteristic length allows for an accurate computation of the spatial derivatives of velocity and, in turn, of WSS, vorticity, ecc..?

 

Regarding the second point. Sotelo et al., 2016 (reference 18) exposed the WSS computations and study different voxel resolutions. We add the references needed in the segmentation and data processing section of the methodology.

 

 

3. In the Introduction, the authors mention helicity among the hemodynamic parameters widely studied in literature. A method for quantifying helical flow in vivo was proposed in Morbiducci et al., 2009, precisely employing 4D flow MRI data of human thoracic aorta. Given the increasing interest of the scientific community in studying helical flow and its beneficial role in human arteries, have the authors considered to include this hemodynamic parameter in their studies?

 

Regarding the last point. We computed several hemodynamic parameters including helicity density. In the statistical analysis, we did not obtain conclusive proof of significant differences between self-gated and non-gated data.  In the four regions of the aorta defined we only found significant differences in the aortic arch in the Wilcoxon signed-rank test.

 

 

Reviewer 2 Report

The authors describe a quantitative comparison between hemodynamic parameters computed from 4D flow Cardiac MRI with and without respiratory self-gating. The hemodynamic parameters (velocity, WSS, etc) are computed based on semiautomatic segmentation, mesh generation and finite element analysis. The comparison is based on a statistical analysis with the use of the Wilcoxon signed-rank test and the Bland-Altman plots.

Remarks.

The hemodynamic parameters computed and studied are not exactly specified.

For example, the blood flow velocity varies essentially during the cardiac cycle. Therefore it is necessary to indicate what a value for velocity the authors present: the peak value or averaged value over the cardiac cycle. The velocity also varies in space, in particular, over a vessel cross-section. Therefore it is necessary to indicate is it an average velocity over a cross-section, or a maximal velocity in the centre of the limen.

The similar remarks are for other hemodynamic parameters: vorticity, kinetic energy, wall shear stress (WSS), etc.  

 

A special care is needed in computation of WSS because of the highly pulsating character of blood flow in aorta. A high temporal gradient causes a high spatial velocity gradient near the wall [Womersley JR. Method for the calculation of velocity, rate flow, and viscous drag in arteries when the pressure gradient is known. Journal of Physiology 1955; 127]. Therefore computation of WSS requires a special boundary layer mesh in vicinity of the vessel wall [Dyedov et al. Variational generation of prismatic boundary-layer meshes for biomedical computing. International Journal for Numerical Methods in Engineering 2009; 79]. A uniform mesh with elements of 1 mm size could be too coarse to resolve correctly the velocity gradient near vessel wall and calculate the WSS. This issue should be discussed and corrected if necessary.

 

The semiautomatic segmentation method employed in the work should be described or a reference on it should be presented.

 

Also it would be better to describe details of application of the Wilcoxon signed-rank test.

 

Minor remark. Line 61: comma is missed after the word “helicity”.

 

The decision about publication can be made after all the indicated remarks have been addressed.

Author Response

The authors describe a quantitative comparison between hemodynamic parameters computed from 4D flow Cardiac MRI with and without respiratory self-gating. The hemodynamic parameters (velocity, WSS, etc) are computed based on semiautomatic segmentation, mesh generation and finite element analysis. The comparison is based on a statistical analysis with the use of the Wilcoxon signed-rank test and the Bland-Altman plots.

 

Thank you for your comments.

 

Remarks.

 

1- The hemodynamic parameters computed and studied are not exactly specified. For example, the blood flow velocity varies essentially during the cardiac cycle. Therefore it is necessary to indicate what a value for velocity the authors present: the peak value or averaged value over the cardiac cycle. The velocity also varies in space, in particular, over a vessel cross-section. Therefore it is necessary to indicate is it an average velocity over a cross-section, or a maximal velocity in the centre of the limen. The similar remarks are for other hemodynamic parameters: vorticity, kinetic energy, wall shear stress (WSS), etc.

 

Regarding the first point, we add references to the computation of hemodynamic parameters previously developed using a least-square projection method based on finite elements. Those references contain the detailed derivation of the hemodynamic parameters analyzed.  To perform the statistical analysis of the hemodynamic parameters presented we used the peak systole cardiac cycle information, and we computed the mean of the measurements in each region of the aorta.

 

 

2- A special care is needed in computation of WSS because of the highly pulsating character of blood flow in aorta. A high temporal gradient causes a high spatial velocity gradient near the wall [Womersley JR. Method for the calculation of velocity, rate flow, and viscous drag in arteries when the pressure gradient is known. Journal of Physiology 1955; 127]. Therefore computation of WSS requires a special boundary layer mesh in vicinity of the vessel wall [Dyedov et al. Variational generation of prismatic boundary layer meshes for biomedical computing. International Journal for Numerical Methods in Engineering 2009; 79]. A uniform mesh with elements of 1 mm size could be too coarse to resolve correctly the velocity gradient near vessel wall and calculate the WSS. This issue should be discussed and corrected if necessary.

 

Sotelo et al., 2016 (reference 18) exposed the WSS computations and study different voxel resolutions. We add the references needed in the segmentation and data processing section of the methodology.

 

 

The semiautomatic segmentation method employed in the work should be described or a reference on it should be presented.

 

Regarding the semiautomatic segmentation process, we used the in-house MATLAB toolbox developed by Sotelo et al., 2019 (reference 17).

 

 

Also it would be better to describe details of application of the Wilcoxon signed-rank test.

 

To perform the Wilcoxon signed-rank test we used the statistical analysis software SPSS. We report this in the Statistical Analysis section of the methodology in the manuscript.

 

 

Minor remark. Line 61: comma is missed after the word “helicity”.

The decision about publication can be made after all the indicated remarks have been addressed.

 

 

Round 2

Reviewer 1 Report

The authors adequately addressed the raised issues. I recommend the paper for publication.

Author Response

The authors adequately addressed the raised issues. I recommend the paper for publication.

 

Thank you for your recommendation for publication of our paper. The previous comments help us to improve our paper to meet the standards for publication in the special issue of Applied Sciences, Biomedical Imaging Technologies for Cardiovascular Disease.

Reviewer 2 Report

The authors have partly addressed my previous remarks. They have introduced the important references. Nevertheless they did not resolve the ambiguity in indicated parameters. Most hemodynamic parameters are varying in time and space.

For example all parameters which spatial distribution is indicated in Figure 1 vary in time. Are they averaged or peak values?

As for parameters indicated in Table 1, the authors should indicate if they are dealing with values averaged over a heartbeat or peak values, also averaged over a certain cross-section or maximum in a lumen (velocity, vorticity, etc) or maximum in a section. The readers should know exactly which parameters are indicated without reading previous papers.

Therefore the paper needs more revision.

Author Response

The authors have partly addressed my previous remarks. They have introduced the important references. Nevertheless they did not resolve the ambiguity in indicated parameters. Most hemodynamic parameters are varying in time and space.

For example all parameters which spatial distribution is indicated in Figure 1 vary in time. Are they averaged or peak values?

As for parameters indicated in Table 1, the authors should indicate if they are dealing with values averaged over a heartbeat or peak values, also averaged over a certain cross-section or maximum in a lumen (velocity, vorticity, etc) or maximum in a section. The readers should know exactly which parameters are indicated without reading previous papers.

Therefore the paper needs more revision.

 

Thank you for your comments. In the methodology we clarify that all parameters were obtained at a representative peak systolic cardiac phase (line 110), and in the description of figure 1 and table 2, we clarify that hemodynamic parameters were measured at a representative peak systole cardiac phase.

We appreciate your comments, and we hope that our paper meets the standards for publication in the special issue of Applied Sciences, Biomedical Imaging Technologies for Cardiovascular Disease.