Whole-Heart Assessment of Turbulent Kinetic Energy in the Repaired Tetralogy of Fallot
Round 1
Reviewer 1 Report
Approximately 10% of congenital heart diseases include Tetralogy of Fallot. Turbulent kinetic energy (TKE), as derived from 4D-flow magnetic resonance imaging, has been used to characterize abnormal heart hemodynamic in CHD.
Comments:
1) a child has to lie in an MRI tube for 10 minutes for TKE measurements. What fundamentally new do you get with this that cannot be seen with an ultrasound, spending much less time and not traumatising the psyche of a child who has to lie in a tube for 10 minutes, with the child lying down normally and not moving around. What is the accuracy of such measurements when baby is moving?
2) What conclusions of the work can be used in medical practice specifically? Do doctors know very well that there will be turbulent and stenotic flow after any operation for Tetralogy of Fallot? It is much easier to assess right ventricular strein and function on ultrasound.
3) Fig. 2 (b), (c), (d) please make bigger in size
4) Please write more on technical details. What is the principles of your Matlab subroutines and codes.
Author Response
Reviewer 1
Approximately 10% of congenital heart diseases include Tetralogy of Fallot. Turbulent kinetic energy (TKE), as derived from 4D-flow magnetic resonance imaging, has been used to characterize abnormal heart hemodynamic in CHD.
R1-C1: A child has to lie in an MRI tube for 10 minutes for TKE measurements. What fundamentally new do you get with this that cannot be seen with an ultrasound, spending much less time and not traumatising the psyche of a child who has to lie in a tube for 10 minutes, with the child lying down normally and not moving around. What is the accuracy of such measurements when baby is moving?
We appreciate the reviewer to this comment as it allows us to clarify some aspects from the 4D-flow and TKE calculation. First, we have added in the introduction some detail in the use of Doppler ultrasound as primary imaging modality, we also commented in the limitations that ultrasound has to properly quantify and characterize flow and turbulence measurements. Second, we want to clarify that the cohort employed in this study are adults. However, we have included some comments in the discussion for the specific use in both pediatric and adult populations. We discussed the importance of a reduced scan time to reduce emotional stress. We have highlighted the effect of motion during acquisition as source of error.
Introduction was modified as follows:
“Transthoracic echocardiography is the first line imaging modality to assess cardiovascular diseases given its easy accessibility, low cost, and safety.1,2 However, it may limited by poor acoustic windows, beam alignment, operator-dependence, and inaccuracy for quantifying regurgitant lesion 3–5. Currently, standard cardiac magnetic resonance (CMR) is the gold standard for the monitoring of heart function and remodeling in patients with rTOF 6. […] In addition, noninvasive Doppler methods allow us to measure adverse velocity fluctuations, but they are also limited in one direction due to the restriction of the ultrasound beam and have an elevated user variability 7,8.”
Discussion section was modified as follows:
“In the current study we derived TKE from a whole-heart 4D-flow MRI acquisition in adult patients. It is important to remark that this kind of acquisition can also be performed in pediatric patients by adjusting the spatial resolution, and velocity encoding parameters 9. In adults a spatial resolution of 2.5 - 3.0 mm3 is recommended, in pediatric 1.5 mm3, and neonates 0.75 – 1.0 mm3. The advantages of 4D-flow MRI in neonatal and pediatric populations over standard 2D phase-contrast are well documented 10,11. The advances in acceleration techniques have reached clinically acceptable scan times, the possibility of free-breathing protocols, and the use of feed and wrap has become more practical to diminish emotional stress 12. However, the use of sedation and anesthesia is still clinical routine for facilitating the exam 13,14. In both adult and pediatric patients, motion artifacts can impact the accuracy of the 4D-flow acquisition. Respiratory motion is usually well managed by using a respiratory navigator, as we used in our study cohort. Novel free-breathing methods manage motion effects within the image acquisition framework 15,16. However, large motion effects cannot be effectively corrected”
R1-C2: What conclusions of the work can be used in medical practice specifically? Do doctors know very well that there will be turbulent and stenotic flow after any operation for Tetralogy of Fallot? It is much easier to assess right ventricular strain and function on ultrasound.
The added value of TKE in clinical practice is still under investigation, particularly in the context of rTOF where limited work has been performed. As we presented in the discussion section, elevated TKE can be observed in the RV in presence of significant PR which is also associated with RV remodeling. The latter was also observed in our study, not only at the time of regurgitation but throughout the cardiac cycle. We have added a short comment highlighting the clinical impact of this finding in the discussion section. We also contextualize with respect to strain and function as given by ultrasound and MRI. It is important to remark the different kind of information that we are obtaining from function and strain in comparison with 3D flow imaging.
Following changes were included:
“In particular, Fredriksson et al. highlighted the potential clinical value of TKE in the development of late complications after TOF repair and the importance of follow-up 17. TKE comes to provide additional information beyond the heart function and strain characterization that can be achieve with standard-of-care ultrasound and MRI. It is reasonable to consider that these basic metrics can support decision making, but they could be other factors besides RV volume and deformation rate that can contribute to an adverse outcome. Flow-derived metrics, as TKE, could provide major understanding of the 3D intra-cardiac hemodynamics and local alterations within the blood flow. It is important to remark that 3D hemodynamics are not fully characterized using standard ultrasound or MRI.”
R1-C3: Fig. 2 (b), (c), (d) please make bigger in size
We modified Figure 2 by enlarging and re-organizing all elements.
R1-C4: Please write more on technical details. What is the principles of your Matlab subroutines and codes.
We modified the entire section 2.5 to include technical details. Most tools have been used in previous publications and have been updated for version compatibility. TKE tool calculation algorithm was shared by Dyverfeldt et al. Our TKE tool basically integrate multiple existing resources to facilitate the analysis and reporting of in-vivo data.
Changes were as follow:
“As shown in Figure 1, after acquisition 4D-flow data were corrected for eddy currents, i.e., linear phase drift according to static tissues, noise masking, and velocity aliasing using “Velomap_tool”, a Matlab tool developed by Bock et al. in 2007 and which is broadly used by the flow MRI community 18. Our in-house tool “4D-Flow Analysis Tool” was developed in MATLAB 2020b (Mathworks, Natick, MA, USA) and integrates the Velomap tool for data pre-processing. After data correction, an individual phase-contrast magnetic resonance angiogram (PC-MRA) throughout the entire cardiac cycle was used to segment the following vessels: Aorta, Pulmonary Artery (PA), Left Atria (LA), Left Ventricle (LV), Right Atria (RA) and Right Ventricle (RV). The PC-MRA is given by , where is the magnitude image, is the spatial location within the volume, is the velocity-encoded image with j representing the velocity encoding direction in image coordinates (x, y, z), and i is the measured time frame in the cardiac cycle. This segmentation approach has been previously reported 19,20. Furthermore, each vessel was divided into several regional volumes for facilitating data analysis. The Aorta was divided into the following 4 segments: aortic root, arch, ascending aorta, and descending aorta. The PA was divided into the following 3 segments: left pulmonary artery (LPA), right pulmonary artery (RPA), and mid pulmonary artery (MPA). Lastly, the LA, LV, RA, and RV were divided into the following 2 segments: inferior, and superior.
After segmentation, an in-house Matlab tool “YYC 4D Flow TKE Tool” was used to calculate TKE throughout the entire cardiac cycle for each segmentation as demonstrated in Figure 2. The magnitude of the individual velocity-encoding directions was reconstructed to compute TKE. The TKE is defined as , where is the fluid density, and is the fluctuation intensity in the three orthogonal directions 21. TKE has been validated in-vitro and in-vivo by several groups 21–25. Our implementation is based on the original code from Dyverfeldt et al. 21. Our TKE tool facilitates the integration of the individual segmentations and TKE calculation to generate standardized analysis reports using regions of interest and templates.”
Author Response File: Author Response.pdf
Reviewer 2 Report
In this manuscript, the authors compare the turbulent kinetic energy (TKE) in the heart and large vessels of 17 patients with repaired Tetralogy of Fallot (rToF) with the TKE in 18 healthy subjects. They use in-house tools to calculate TKE and standardize the results for subsequent analyses. The authors find interesting results regarding the (statistically significant) differences between the max, min, mean, etc. TKE values in patients and healthy subjects. Results show that TKE could indeed be an interesting parameter that should be analyzed when monitoring a patient with rToF.
Before recommending the acceptance of this manuscript, I would like to discuss the following 4 points with the authors.
-The authors use in their study in-house tools and codes to calculate TKE. Are these tools and codes validated? In other words, the results provided by these tools should have been compared against reality (if possible). Are the results provided by these tools accurate enough?
-At the moment, the authors have compared TKE values in patients with TKE values in healthy subjects. Have they associated these TKE values with the onset of any complications taking place in patients with rToF? I have the feeling that this study would be more significant if actual TKE values (max, min, mean, etc.) were associated with specific complications.
-Have the authors considered other patients whose heart and large vessel hemodynamics are altered? For example, Fontan patients.
-Would computational fuid dynamics (CFD) simulation-based calculation of TKE be a potential tool for the analyisis of hemodynamics in rToF patients? Other modeling issues would arise as a result of running a CFD simulation (e.g., turbulence modeling, vessel wall-hemodynamics coupling, etc.), but it could be used maybe to optimize heart and vessel geometry to improve TKE-based hemodynamic indices.
Additionally, there are some minor comments below:
-Line 111: Reference 19 is the same as reference 18, therefore reference 19 should be removed.
-Line 131: It should say LVESV instead of LVSEV.
-Line 167: It would be helpful to complete the explanation about the standard deviation. Por example: "standard deviation of the component of the velocity vector at the i-direction" or something similar.
-Line 188: It might be better to say "statistical significance" than simply "significance".
-Line 190: What is the meaning of "indexed"? Is it the variable per unit BSA?
-Line 193: It might be better to say "statistical significance" than simply "significance".
-Table 1: Units of diastolic blood pressure should say "mmHg" instead of "mmHG".
-Lines 223 and 224: Why is "(LVEF)" after Indexted Left Ventricular Mass and after Left Ventricular Mass? I think they should not be there, otherwise I might be missing something.
-Line 338: Did the authors mean "simplify"? I guess this sentence is not correct.
-Line 339: Did the authors mean "not"?
-Line 340: What does AHA stand for? Is this the American Heart Association?
Author Response
Reviewer 2
In this manuscript, the authors compare the turbulent kinetic energy (TKE) in the heart and large vessels of 17 patients with repaired Tetralogy of Fallot (rToF) with the TKE in 18 healthy subjects. They use in-house tools to calculate TKE and standardize the results for subsequent analyses. The authors find interesting results regarding the (statistically significant) differences between the max, min, mean, etc. TKE values in patients and healthy subjects. Results show that TKE could indeed be an interesting parameter that should be analyzed when monitoring a patient with rToF.
Before recommending the acceptance of this manuscript, I would like to discuss the following 4 points with the authors.
R2-C1: The authors use in their study in-house tools and codes to calculate TKE. Are these tools and codes validated? In other words, the results provided by these tools should have been compared against reality (if possible). Are the results provided by these tools accurate enough?
We appreciate this suggestion. Please see response R1-C4. We briefly gave more details on the implementation of the tools. Most of these tools have been widely used and validated. Our TKE tool, in particular, facilitates the integration of individual segmentations for standardized analysis and reporting. We have also added a reference to a recent review article that effectively summarize the limitations of TKE calculation with MRI.
Discussion modification is as follows:
“As reported by Dillinger et al., the level of intra-voxel underestimation will depend on the turbulence level and velocity encoding strategy 26. Single encoding gradient can vary by 20%. A Lagrangian velocity spectrum on a voxel-by-voxel basis may correct the latter.”
R2-C2: At the moment, the authors have compared TKE values in patients with TKE values in healthy subjects. Have they associated these TKE values with the onset of any complications taking place in patients with rToF? I have the feeling that this study would be more significant if actual TKE values (max, min, mean, etc.) were associated with specific complications.
We appreciate this comment. This study was designed as a pilot to compare TKE with standard clinical parameters. As the reviewer could notice, on this manuscript our lab also reports the development of a TKE analysis tool for facilitating analysis and reporting. The latter is part of the pilot phase before moving on a more comprehensive clinical study. We agree that a more significant clinical value can be achieved by investigating the association of TKE with rTOF complications. We reviewed our patient cohort; the number and type of complications are small to perform a valid statistical analysis. In consideration of the importance of your remark, we have included two comments in the discussion section as follows:
“The association of TKE with the onset of any rTOF complications must be investigated to understand the TKE role in adverse outcomes. […] Patient sample size also limited the ability to investigate specific complications (e.g. pulmonary regurgitation).”
R2-C3: Have the authors considered other patients whose heart and large vessel hemodynamics are altered? For example, Fontan patients.
For instance, our TKE analysis was solely centered on rTOF. Fontan patients referred to our center typically don’t get a 4D-flow acquisition. However, we think Fontan patients can present conditions on which TKE could give novel hemodynamic information. A good example it a recent numerical study from Grunwald et al. on which vortex formation and TKE demonstrated a deteriorated cardiac performance while EF remained normal.
We have included a comment in the discussion as follows:
“Our study only considered rTOF patients for the investigation of TKE. However, other congenital diseases could benefit from the characterization of abnormal TKE. A recent numerical study demonstrated that lower vortex formation and lower TKE can identify deteriorated intra-cardiac performance in Fontan patients while ejection fraction did not, whereas the latter still needs to be demonstrated in-vivo 27.”
R2-C4: Would computational fluid dynamics (CFD) simulation-based calculation of TKE be a potential tool for the analysis of hemodynamics in rToF patients? Other modeling issues would arise as a result of running a CFD simulation (e.g., turbulence modeling, vessel wall-hemodynamics coupling, etc.), but it could be used maybe to optimize heart and vessel geometry to improve TKE-based hemodynamic indices.
This is an interesting comment. In the past, CFD has been used as a reference for TKE measurements derived from 4D-flow MRI. The group from Linkoping has several works on this aspect. In particular, a study from Petersson et al. MRM 2016 is often referred for experimental validation of TKE. There are a small number of studies assessing rTOF using CFD. In particular, the study from Loke et al. reported a framework for computational modeling using cardiac MRI images and 4D-flow for patients with rTOF which included the calculation of kinetic energy, TKE and vorticity. Most groups have used CDF for improving the calculation of wall shear stress, as 4D-flow spatial and temporal resolution can underestimate this measurement. For TKE, CFD can facilitate the understanding of the turbulence cascade at Kolmogorov microscales. The 4D-flow in-vivo acquisition mostly reflets the macro scale of turbulence, as we acquire the data with a 2-3 mm3 resolution. However, commercial CDF approaches could be computationally demanding and time consuming. Novel methods using 3D Lattice Boltzmann could be more effective for clinical applications.
We have included these aspects in the discussion section as follows:
“The use of computational fluid dynamics (CFD) to complement and/or validate 4D-flow derived calculations is not new. TKE, as it is calculated our study, has been validated using experimental models and CFD reporting good agreement between MRI, models and simulations 28. Some CFD studies have evaluated the influence of the pulmonary artery bifurcation angle, pressure distribution, flow patterns 29,30. In particular, Loke et al. reported a framework for computational modeling using cardiac MRI images and 4D-flow MRI in rTOF patients and included the calculation of kinetic energy, vorticity and TKE in rTOF 31. However, these studies were most exploratory with a limited number of subjects. Furthermore, CFD has also been used to improve the calculation of wall shear stress, as 4D-flow spatial and temporal resolution can underestimate this measurement 32–34. Casas et al. used CFD to assess the impact of spatial resolution and reported that TKE estimates were accurate, minorly impacted by resolution while viscous energy dissipation was underestimated and showed resolution dependance 35. However, 4D-flow in-vivo acquisition mostly reflects the macro scale of turbulence, as data are acquired with a 2-3 mm3 resolution 36. CFD can facilitate the TKE understanding of the turbulence cascade at the Kolmogorov micro scale. Something that cannot be achieve in-vivo. Exploring the hemodynamic data with CFD in such scale can be computationally demanding and time consuming. Some level of feasibility can be achieved using novel 3D Lattice Boltzmann methods which have proved to be more efficient for clinical applications 37.”
R2-C5: Line 111: Reference 19 is the same as reference 18, therefore reference 19 should be removed.
Corrected.
R2-C6: Line 131: It should say LVESV instead of LVSEV.
Corrected.
R2-C7: Line 167: It would be helpful to complete the explanation about the standard deviation. Por example: "standard deviation of the component of the velocity vector at the i-direction" or something similar.
Corrected.
R2-C8: Line 188: It might be better to say "statistical significance" than simply "significance".
Corrected.
R2-C9: Line 190: What is the meaning of "indexed"? Is it the variable per unit BSA?
Yes, it is BSA indexation. We have clarified in the manuscript.
R2-C10: Line 193: It might be better to say "statistical significance" than simply "significance".
Corrected.
R2-C11: Table 1: Units of diastolic blood pressure should say "mmHg" instead of "mmHG".
Corrected.
R2-C12: Lines 223 and 224: Why is "(LVEF)" after Indexted Left Ventricular Mass and after Left Ventricular Mass? I think they should not be there, otherwise I might be missing something.
We have modified the LVEF location.
R2-C13: Line 338: Did the authors mean "simplify"? I guess this sentence is not correct.
Corrected.
R2-C14: Line 339: Did the authors mean "not"?
Corrected.
R2-C15: Line 340: What does AHA stand for? Is this the American Heart Association?
Yes, now defined in the document.
Author Response File: Author Response.pdf