Atlas-Based Evaluation of Hemodynamic in Ascending Thoracic Aortic Aneurysms

Round 1

Reviewer 1 Report

I believe this is a very interesting work wich sets the path towards useful prognostic tools for ATAA patients. The document is very clear and well presented. Both the methods and results are very well described and analyzed. The use of advanced parametrisation tools such as SSM is very well aliged with the use of reduced order modeling and machine learning techniques in biomechanics, where the parametrization of the geometries is a key limiting factor. I have a minor question/comment, related to the interpretability of the results. It is clear that the identified modes correlate with hemodynamic alterations. The authors also identify relationships between those modes with specific features of the template geometry, such as aortic size, tortuosity or aortic valve geometries. Yet, this relationship between modes and features of the geometry are identified visually from the variation of the dominant shape modes (see for example  Figure 5). Yet, this is more a qualitative relationship as  the modes themselves come from PCA analysis and therefore do not have a specific physical meaning. The authors also mention that they would like to use Deep Learning in the future, to correlate such modes with biomechanical responses. Combining the non-physical meaning of PCA modes with a black box type model using deep learning, this would end up resulting in a non-interpretable tool, which can pose barriers to their use in clinics. Have the authors thought about this limitation? Is there a way to correlate those modes with physics based metrics that describe the geometry of the aorta? Would there be a way to predict biomechanics without resorting into black box type tools such as deep learning? 

Author Response

Reviewer #1

We thank the reviewer for his or her valuable comments. We have taken these comments into careful consideration when preparing the revised manuscript and feel that the critiques led directly to an improved submission. We hope that the reviewer agrees. All changes made to the document were highlighted in yellow.

Major Comment

I believe this is a very interesting work wich sets the path towards useful prognostic tools for ATAA patients. The document is very clear and well presented. Both the methods and results are very well described and analyzed. The use of advanced parametrisation tools such as SSM is very well aliged with the use of reduced order modeling and machine learning techniques in biomechanics, where the parametrization of the geometries is a key limiting factor.

 

I have a minor question/comment, related to the interpretability of the results. It is clear that the identified modes correlate with hemodynamic alterations. The authors also identify relationships between those modes with specific features of the template geometry, such as aortic size, tortuosity or aortic valve geometries. Yet, this relationship between modes and features of the geometry are identified visually from the variation of the dominant shape modes (see for example  Figure 5). Yet, this is more a qualitative relationship as  the modes themselves come from PCA analysis and therefore do not have a specific physical meaning. The authors also mention that they would like to use Deep Learning in the future, to correlate such modes with biomechanical responses. Combining the non-physical meaning of PCA modes with a black box type model using deep learning, this would end up resulting in a non-interpretable tool, which can pose barriers to their use in clinics. Have the authors thought about this limitation? Is there a way to correlate those modes with physics based metrics that describe the geometry of the aorta? Would there be a way to predict biomechanics without resorting into black box type tools such as deep learning?

Reply: We are happy to hear that the Reviewer has appreciated the present study. With regards to the minor question/comment, we specify that the PCA output can be visualized in two way: 1) the geometrical shape for visual inspection (as Figure 5) and 2) shape vector as numbers representing the variance of each patient from the mean shape. While the shape mode can be used for visually inspecting the changes in the aortic geometry, the shape vector can be used for quantitative correlation with clinical and functional data. This was the goal of our previous investigation where the relationship between morphological parameters (ie, aortic size, tortuosity) and wall shear stress was investigated. Thus, the PCA output can be also a quantitative output that can be used for correlation analyses. We agree with the Reviewer that patient-specific analysis of ATAA biomechanics are indeed better with respect to the deep learning (as a black type box). The following sentences was added in the Study Limitation section:

“As reported by Cosentino et al. [10], shape vectors for each patients with ATAAs can be obtained from PCA and then used to assess the correlation between morphological fea-tures (ie, aortic size and tortuosity) and biomechanical functional parameters (ie, WSS). In this study, the SSM was adopted to generate a virtual patient cohort and assess the hemodynamic in borderline ATAA shapes.”

 

Reviewer 2 Report

Review report of the manuscript applsci-1502921 by Catalano et al.

The authors around C. Catalano carry out an atlas-based disease assessment of the computational flow dynamic of the dilated ascending aorta. A shape modeling was created to parametrize the aortic shape and its variability. For each patient group, the mean shape was deformed upon the first three principal shape modes, and then the relationship between shape and computational flow parameters was investigated to quantify the hemodynamic of the aneurysmal geometries.

General Comments

The work is well organized and interesting. In my opinion it makes a contribution in the field and would be interesting for the readers. However, several major concerns should be addressed before the manuscript can be accepted for publication.

Specific comments

• The introduction is well redacted but I personally feel a lack of inclusion of additional references. Only 8 studies are cited. However, in my opinion it is necessary to add many other studies. As an example, there are several additional studies that consider the vessel anatomical and geometrical features a part this cited by the authors (only 3...). The geometrical features and parameters have been related to the human hemodynamics for the aorta and even for different vessels. These patient specific and parametric models should be added and discussed.
• Paragraph 2.3: The SSM approach is based on the surface. Did the author consider the vessel centerline as well for comparing the geometrical features? In the literature, this analysis has been performed by different authors.
• Which criterion has been used for deforming the shape modes? Do the authors compared the deformed shape with physiological shapes to ensure the the analyzed variations are reasonable?
• Paragraph 2.4: why is laminar flow used? The flow in the aorta is usually considered as turbulent and in particular, k-w SST model has been demonstrated to be the best model capable of providing the aortic flow features.... 
• Are the boundary conditions applied as flat profiles on the extensions surfaces? Please add and discute.
• How is the aortic valve reconstructed? Fully open surface during the entire duration of the heart cycle? Please add and discute.

• 'The peak systolic flow velocity has an helical flow..' What about helicity? How do you consider this In the modeling?

• In the Discussion, the authors state: 'The atlas-based disease evaluation here proposed has demonstrated the ability to reveal the hemodynamic impairment induced by specific aneurysm and aortic valve features'. However, it is not clear how the valve is considered and which conditions actually reflect the considered shape variation, as there is no comparison with patient-specific aneurysm models..

• Lines 225-228: Please add more references. There are much more studies stating and studying similar topics that is worths mention.
• Lines 270-272: Please discute the tortuosity with studies in the literature as well.

Technical Comments

Some typos and style errors will be listed below. However, please have a look on the entire manuscript.

• Line 63: anerusym
• Lines 121-122: 'laminar-flow fluid with non-Newtonian viscosity'. Please rephrase as 'laminar non Newtonian flow', or something similar.
• Line 165: Figure 3. and 4
• Line 169: Other Modes

Author Response

Reviewer #2

We thank the reviewer for his or her valuable comments. We have taken these comments into careful consideration when preparing the revised manuscript and feel that the critiques led directly to an improved submission. We hope that the reviewer agrees. All changes made to the document were highlighted in yellow.

General Comments

The authors around C. Catalano carry out an atlas-based disease assessment of the computational flow dynamic of the dilated ascending aorta. A shape modeling was created to parametrize the aortic shape and its variability. For each patient group, the mean shape was deformed upon the first three principal shape modes, and then the relationship between shape and computational flow parameters was investigated to quantify the hemodynamic of the aneurysmal geometries.

The work is well organized and interesting. In my opinion it makes a contribution in the field and would be interesting for the readers. However, several major concerns should be addressed before the manuscript can be accepted for publication.

Reply: We are glad to know that the Reviewer believes that this manuscripts can contribute to the field of biomechanics of ATAAs. We did our best to clarify all comments and suggestions arose by Reviewer.

 

Specific Comments

The introduction is well redacted but I personally feel a lack of inclusion of additional references. Only 8 studies are cited. However, in my opinion it is necessary to add many other studies. As an example, there are several additional studies that consider the vessel anatomical and geometrical features a part this cited by the authors (only 3...). The geometrical features and parameters have been related to the human hemodynamics for the aorta and even for different vessels. These patient specific and parametric models should be added and discussed.

Reply: additional references describing the link between the function and shape as well as the importance of computational analysis in ATAAs were added. The following paragraph was added in the Introduction section:

“The understanding of the mechanistic link between the shape and function can reveal important insight not only on the aneurysm disease, but also for the development of novel risk metrics not based on the aortic size [12, 13]. According to Laplace’s law, the large ATAA wall experiences a greater level of intramural stress than a non-aneurysmal aorta [14]. The intramural stress was demonstrated to provide better predictive capability than the aortic size [15, 16], even in the classic versus the bovine aortic arch [17]. Com-putational flow analyses have also evinced different patterns of fluid shear forces and eccentric flow in ATAAs with different morphological features [18, 19] and valve phe-notypes [20]. In the setting of the aortic arch coartaction, SSM was used to asses the link between shape features and biomechanical descriptors [9] as well as detect shape cluster of patients at high risk of complications [21]. “    

 

Paragraph 2.3: The SSM approach is based on the surface. Did the author consider the vessel centerline as well for comparing the geometrical features? In the literature, this analysis has been performed by different authors.

Reply: In our previous investigations (see Cosentino et al [10]), we investigated the importance of geometrical features of each ATAA with the aneurysm function. Instead of using the centerline, we evaluated the ascending aortic curvature and tortuosity as indicators of the 3D nature of ATAA. The goal of the present study was to first develop the shape atlas and then assess the computational flow in borderline cases from a 3D point of view rather than as global value for each patient. The following sentence was added in the Study Limitation section:

“As reported by Cosentino et al. [10], shape vectors for each patients with ATAAs can be obtained from PCA and then used to assess the correlation between morphological fea-tures (ie, aortic size and tortuosity) and biomechanical functional parameters (ie, WSS). In this study, the SSM was adopted to generate a virtual patient cohort and assess the hemodynamic in borderline ATAA shapes”

 

Which criterion has been used for deforming the shape modes? Do the authors compared the deformed shape with physiological shapes to ensure the the analyzed variations are reasonable?

Reply: Indeed, the shape modes were not deformed at highest boundary of 3 SD to avoid unrealistic aneurysm shapes. However, not any physiological data was used to determine the final deformed shape (ie, value of SD) but rather the SSM was deformed upon the arbitrary shape deformation of 1.5 SD. Nothing was added in the text if permitted by the Reviewer as this is described in Section 2.4. 

 

Paragraph 2.4: why is laminar flow used? The flow in the aorta is usually considered as turbulent and in particular, k-w SST model has been demonstrated to be the best model capable of providing the aortic flow features.... 

Reply: There are many studies in the literature suggesting that a laminar flow could be a realistic assumption. Indeed, laminar flow usually occurs in large vessels due to the low mean flow velocity. For the mean ATAA shape, we observed a Reynolds number of 1434 at the mid-height of the ascending aorta. The following sentences were added int the Method Section to support this assumption.

“The blood flow was assumed to be laminar since this usually occurs in large vessels due to the low mean flow velocity. For the BAV ATAA template model, we found an average Reynolds number of 1434 at mid-height of the ascending aorta.”

 

 

Are the boundary conditions applied as flat profiles on the extensions surfaces? Please add and discute.

Reply: Parabolic profile are used as boundary conditions at the inlet. The sentence was rewritten as a follow to clarify this aspect:

“For the inflow, all models were set with a representative parabolic flow profile with fundamental frequency of 1 Hz and peak velocity of 1.2 m/s as described previously [28].”

 

 

How is the aortic valve reconstructed? Fully open surface during the entire duration of the heart cycle? Please add and discute.

Reply: The valve is fully open for the entire cardiac cycle. At diastole, the inflow velocity is set to zero to simulate the closed valve. Indeed, we recognize that this is a limitation so that the following sentence was added in the Study Limitation section:

“As a standard computational fluid dynamic approach was used, the aortic valve leaflets were rigid as segmented at fully opened shape from CT scan at systole. In future studies, for realistic flow simulation, fluid-solid interaction will be used to simulate the motion of the aortic valve during the cardiac beating.”

 

 

'The peak systolic flow velocity has an helical flow..' What about helicity? How do you consider this In the modeling?

Reply: In our previous study (Pasta et al EJVS 2018), we observed different values of the helicity between BAV ATAA and TAV ATAAs but without statistical difference between the two groups. Nothing was added in the text if permitted by the Reviewer.

 

In the Discussion, the authors state: 'The atlas-based disease evaluation here proposed has demonstrated the ability to reveal the hemodynamic impairment induced by specific aneurysm and aortic valve features'. However, it is not clear how the valve is considered and which conditions actually reflect the considered shape variation, as there is no comparison with patient-specific aneurysm models..

Reply: Here, we was referring that the study included patients with different valve phenotype. The sentence was rewritten to specify this aspect as a follow:

“The atlas-based disease evaluation here proposed has demonstrated the ability to reveal the hemodynamic impairment induced by specific aneurysm features and aortic valve phenotypes-, (ie, the BAV versus TAV phenotype).”

 

 

Lines 225-228: Please add more references. There are much more studies stating and studying similar topics that is worths mention.

Reply: Please see the reply to the first comment where more reference were added in the study.

 

 

Lines 270-272: Please discute the tortuosity with studies in the literature as well.

Reply: Please see the reply to the second comment where we discussed the importance of tortuosity as done in Cosentino et al.

 

 

Technical Comments

Some typos and style errors will be listed below. However, please have a look on the entire manuscript.

Line 63: anerusym

Reply: fixed

 

 

Lines 121-122: 'laminar-flow fluid with non-Newtonian viscosity'. Please rephrase as 'laminar non Newtonian flow', or something similar.

Reply: fixed

 

 

Line 165: Figure 3. and 4

Reply: fixed

 

 

Line 169: Other Modes

Reply: fixed

 

Reviewer 3 Report

Please, find the attached file

Comments for author File: Comments.pdf

Author Response

Reviewer #3

We thank the reviewer for his or her valuable comments. We have taken these comments into careful consideration when preparing the revised manuscript and feel that the critiques led directly to an improved submission. We hope that the reviewer agrees. All changes made to the document were highlighted in yellow.

 

Major Comment

In this paper, the variability of the shapes of the aortic systems, variability based on atlases of experimental data, is considered by means of statistical methods, identifying different forms taken as a reference (modes). Therefore, for each mode a Computational Fluid Flow (CFD) code, commonly used by the international scientific community was applied (Fluent). The article is interesting and, in my opinion, of concrete usefulness and therefore worthy of being published. However, precisely because of the interest it could find in the scientific community (in particular in the mathematical-medical one) it is necessary that the following details should be specified:

 

The authors mentioned, in line 226, the approach of FSI (fluid structure interaction) which would highlight how the deformability of the aortic duct could affect blood speed and pressure; however they did not sufficiently discuss how this circumstance would impact on the results of their work;

Reply: As the aorta enlarge, the vessel becomes stiff so that the assumption of rigid aortic wall is reasonable. Two studies have demonstrated comparing CFD and FSI techniques that there is no changes in the prediction of WSS for vessel diameter changes between systole and diastole (Reymond et al) and for ATAAs (Mendez et al). The following sentence was added in the Study Limitation section:

“In a different way, the ATAA leads to stiffening of the aortic wall so that the assumption of a rigid aortic wall is reasonable as reported in numerical comparison between com-putational fluid dynamic and fluid-solid interaction techniques [43, 44].”

 

A laminar flow regime was assumed; the authors should justify this assumption, also by calculating the Reynolds number;

Reply: There are many studies in the literature suggesting that a laminar flow could be a realistic assumption. Indeed, laminar flow usually occurs in large vessels due to the low mean flow velocity. For the mean ATAA shape, we observed a Reynolds number of 1434 at the mid-height of the ascending aorta. The following sentences were added int the Method Section to support this assumption.

“The blood flow was assumed to be laminar since this usually occurs in large vessels due to the low mean flow velocity. For the BAV ATAA template model, we found an average Reynolds number of 1434 at mid-height of the ascending aorta.”

 

 

The fluid dynamic characteristics are influenced not only by the regime (turbulent or laminar), but also by the composition of the fluid and therefore its profile not only in turbulent conditions, but also in laminar conditions, as is reported in the article: Pasculli A. (2018). Viscosity Variability Impact on 2D Laminar and Turbulent Poiseuille Velocity Profiles; Characteristic-Based Split (CBS) Stabilization. 5th International Conference on Mathematics and Computers in Sciences and Industry (MCSI), Corfu, Greece. DOI: 10.1109 / MCSI.2018.00038; ISBN: 978-1-5386-7500-7.

Reply:  We agree with the Reviewer as few experimental studies using particle image velocimetry suggests that near the aortic valve the flow could be turbolent. The following sentence was added in the Methods section to clarify this aspect:

“Though the Reynolds number is low, the composition of the fluid could lead to turbu-lence under laminar flow conditions.”

 

Actually, along non straight path of the aortic pathways, it is reasonable to expect that uneven distributions of the elements with different densities, contained in the blood, will be generated;

Reply: We agree with the Reviewer that the density of mesh elements can lead to different results. The resulting mesh size was determined from grid convergence analysis as previously reported by our group (Rinaudo 2014 PIME H). Nothing was added in the text if permitted by the Reviewer.

 

-the solution of the CFD equations require the application of numerical stabilization algorithms, which one was selected (for example SUPG, CBS as in the article cited, etc.)?

Reply: As stabilization algorithm, we used the SIMPLE algorithm implemented in FLUENT as stated at page 4 line 142. Nothing was added in the text if permitted by the Reviewer.  

 

Please, detail the characteristics of the Carreau model relating to viscosity and the reason for that selection; among others the following papers could be cited:

• Sonia Tabakova, S., Kutev, N., and Radev, S. (2015). Application of the Carreau Viscosity Model to the Oscillatory Flow in Blood Vessels. Conference: 41st International Conference "Applications of Mathematics in Engineering and Economics" AMEE ’15, Vol. AIP Conf.Proc. 1690. DOI: 10.1063 / 1.4936726.
• Khan, M., Sardar, H., Gulzar, M.M., Ali Saleh Alshomrani, A.S. On multiple solutions of non-Newtonian Carreau fluid flow over an inclined shrinking sheet. Results in Physics, 2018, 926-932.

Reply: We thank the Reviewer for suggesting these references. The article published on the journal was added in the reference while the conference abstract was not included. The reason for using the Carraue model is dictated by the fact that this is widely used in biofluid mechanics.

 

Round 2

Reviewer 2 Report

The authors have responded to all the concerns highlighted during the review process so that I am pleased to recommend the manuscript for publication. 

Author Response

We thank the reviewer. No additional comments is required 

Reviewer 3 Report

please see the attached file

Comments for author File: Comments.pdf

Author Response

Reviewer #3

We thank the reviewer for his or her valuable comments. We have taken these comments into careful consideration when preparing the revised manuscript and feel that the critiques led directly to an improved submission. We hope that the reviewer agrees. All changes made to the document were highlighted in yellow.

Query 1: The sentence is contradictory: the fluid can be in laminar conditions or in turbulent conditions. Furthermore, turbulence is characterized by different regimes, such as the laminar-turbulent transitional regime, fully developed, etc.

Reply: the sentence was re-written as a follow:

“Though the Reynolds number is low, the composition of the fluid could lead to different flow regimes such as the laminar-turbulent and the fully developed transitional regimes.”

 

Query 2: I was not referring to a different "mesh density". Actually, blood is a mixture of particles and substances of different densities. Therefore, in the areas where the direction of the fluid undergoes a variation, the consequent acceleration field causes the separation of the denser elements, with consequent modification of the velocity profile, as described, for example among many others, in the paper that I have indicated in the previous revision. Given the importance and delicacy of the topic treated by the authors, the highest rigor in constructing the mathematical models that would help healthcare professionals is required.

At least, some comments and discussion regarding this issue would be mandatory.

Reply: the following sentence was added:

“In the areas of changes in the fluid flow direction, the acceleration field could lead to the separation of the dense part of the blood by modifying the velocity profile”.