A Novel Method for Lung Image Processing Using Complex Networks

Round 1

Reviewer 1 Report

Thank you for your thoughtful responses to the previous review comments. The changes the authors made have improved the quality of the manuscript. I have no further comments.

Author Response

We appreciate your constructive review very much, it has been very valuable to us.

Reviewer 2 Report

-    Paper structure and organization is good for the most part.
-    The literature review is thorough.
-    The methodology and experiments are comprehensive.
-    The mathematical formulation of the proposed method is presented clearly.
-    The dataset used for this study is highly limited. However, the proposed method achieved good performance on the provided dataset.

The Reviewer also suggests providing the codes of this study online (Github repository).

Author Response

Thank you for taking the time to review our manuscript and for your thoughtful response. The code has been uploaded to the following public Github repository: https://github.com/lalasmurf/lungs/

Reviewer 3 Report

This manuscript focuses on lung image processing by using complex networks. There are many defects in it.
1.There is no section 1.1.
2.In line 174, what does "HU intervals from to Lin Li et al. and Maria Paola Belfiore et al. reports (9,10,11)" mean?
3.In line 188, "remover" should be replaced by "remove".
4.According to lines 240 and 241, Fig. 2(c) is combined from Figs. 2(d), 2(e), and 2(f). In fact, Fig. 2(c) seems to contain Figs. 2(e) and 2(f) only.
5.In line 277, "4*0.74" should be replaced by "4x0.74" to match the calculation symbol requirement of this manuscript.
6.In line 287, the equation number "(1)" is necessary.
7.In line 291, "E={E(....}" is strange. The two "E"s are different.
8.Rd" and "D" in line 291 should be respectively replaced by "4" and "50" or "Rd = 4 and D = 50" in line 292 should be removed.
9.In lines 311 and 312, both "Degree distribution the normal sample network" and "Degree distribution the DILD sample network" need a preposition.
10.In lines 323 and 324, both "Degree distribution the normal sample network" and "Degree distribution the DILD sample network" need a preposition.
11.In lines 333 and 334, both "Degree distribution the normal sample network" and "Degree distribution the DILD sample network" need a preposition.
12.In line 347, "Population distributions comparison" should be replaced by "Population distribution comparisons".
13.In line 354, "average degree" should be replaced by "the average degree".
14.In line 356, "average degree" should be replaced by "the average degree".
15.In lines 361 and 362, "a complex-networks based model based on" is strange. "Based" is used twice.
16.In line 374, "fit" should be replaced by "fits".
17.In line 377, "0.74 mm to 1.48 mm" should be replaced by "0.74-mm to 1.48-mm".
18.In line 463, "comparison" should be replaced by "comparisons".
19.In line 406, "comparison" should be replaced by "comparisons".
20.Some equations are necessary for one to understand the proposed method better.
In conclusion, some equations which more clearly explain the proposed method are needed.

Author Response

Comment 1: There is no section 1.1.

Reply 1:  There was a slight inaccuracy there, which we have now corrected, this subsection had accidentally been omitted.

Comment 2. In line 174, what does "HU intervals from to Lin Li et al. and Maria Paola Belfiore et al. reports (9,10,11)"mean?

Reply 2:  Unfortunately, we noticed too late that this (9,10,11) references are written as static text. We uploaded the correct bibliography inserted the proper references.

Comment 3. In line 188, "remover" should be replaced by "remove".

Reply 3: Done. Because of new text insertion (subsection 1.1.) the line 188, now corresponds with line 190         

Comment 4.  According to lines 240 and 241, Fig. 2(c) is combined from Figs. 2(d), 2(e), and 2(f). In fact, Fig. 2(c) seems to contain Figs. 2(e) and 2(f) only.   

Reply 4: This is a valid point, yet here we need to clarify the reason one can see the emphysema pixels in Fig 2(d) with the naked eye, but not the same thing can be said about them in Fig 2(c). First, the type of tissue density (or its equivalent gray tone color) represented in Fig 2(d) is in reality (compared to GGO and Consolidation) much closer to black than to white. This being said, when represented individually, we rendered Fig 2(d) with such a contrast difference that they seem like light greys – with the sole purpose of making them visible. In fact, when put together (or overlapped) with the other two layers (GGO, Consolidation), their color is almost indistinguishable from black and so, it appears as if they don’t even exist at all in Fig 2(c).

Comment 5. In line 277, "4*0.74" should be replaced by "4x0.74" to match the calculation symbol requirement of this manuscript..

Reply 5: Calculation symbol has now been changed to match the requirement. Now on line 279

Comment 6. In line 287, the equation number "(1)" is necessary.

Reply 6: We marked the equation number "(1)" by MDPI proposed format. Now on line 289.

Comment 7. In line 291, "E={E(....}" is strange. The two "E"s are different.

Reply 7: Fixed notation for edges. Now on line 294.

Comment 8. Rd" and "D" in line 291 should be respectively replaced by "4" and "50" or "Rd = 4 and D = 50" in line 292 should be removed.

Reply 8: Fixed now. Replaced Rd and D by "4" and "50" or "Rd = 4 and D = 50" – these are now on line 294

Comment 9. In lines 311 and 312, both "Degree distribution the normal sample network" and "Degree distribution the DILD sample network" need a preposition.

Reply 9:  We added a preposition as follows: “Degree distribution of the normal sample network" and "Degree distribution of the DILD sample network". Now line 314-315

Comment 10. In lines 323 and 324, both "Degree distribution the normal sample network" and "Degree distribution the DILD sample network" need a preposition.

Reply 10: We added a preposition as fallow: “Degree distribution of the normal sample network" and "Degree distribution of the DILD sample network". Now lines 326-327

Comment 11. In lines 333 and 334, both "Degree distribution the normal sample network" and "Degree distribution the DILD sample network" need a preposition.

Reply 11: We added a preposition as follows: “Degree distribution of the normal sample network" and "Degree distribution of the DILD sample network". Now lines 336-337

Comment 12. In line 347, "Population distributions comparison" should be replaced by "Population distribution comparisons".

Reply 12: This change has been made as you correctly suggested.

Comment 13. In line 354, "average degree" should be replaced by "the average degree".

Reply 13: Fixed.

Comment 14. In line 356, "average degree" should be replaced by "the average degree".

Reply 14: Completed

Comment 15. In lines 361 and 362, "a complex-networks based model based on" is strange. "Based" is used twice.

Reply 15: Removed duplicate word.

Comment 16. In line 374, "fit" should be replaced by "fits".

Comment 17. In line 377, "0.74 mm to 1.48 mm" should be replaced by "0.74-mm to 1.48-mm".

Reply 17: Fulfilled this requirement. Line 380

Comment 18. In line 463, "comparison" should be replaced by "comparisons".

Reply 18: Done

Comment 19. In line 406, "comparison" should be replaced by "comparisons".

Reply 19: Done

Comment 20. Some equations are necessary for one to understand the proposed method better. In conclusion, some equations which more clearly explain the proposed method are needed.

Reply 20: Thank you for the equations suggestion. However, we feel that a more theoretical complex networks approach does not suit the audience of this journal as it would lean heavily towards information theory and not on the practical tomography application. Due to your interest, we are already considering creating a new paper (a more technical one) which would include such equations.

Round 2

Reviewer 3 Report

This revised manuscript is ready to be published.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The authors have utilized complex networks to delineate the different densities within a given pixel(s) that might be able to differentiate different ILD states. While this was a very clever idea and I can see the potential of this modeling to potentially help delineate different ILD phenotypes, I have some major concerns about the study as it currently reads:

• Why was only one 65x65 pixel area selected from the 60 patients for analysis when the lung is a large 3D volume? Also, the paper states that the this area was manually selected? I struggle to see the efficacy of this modeling approach when only one small area was chosen from each patient, please provide this modelling approach for the entire 3D volume of the lung. I understand that the complex network approach herein, was designed to probe distributions in overlapping sequelae (i.e. emphysema and consolidations) but this would be a more thorough approach, also by only choosing one pixel area there is no way to know whether your modelling approach is reproducible (both inter and intra-observer variability). 
• Given that only one pixel area was used for each patient is curious and calls into question how long this computational process takes for one small area of the lung? Does this technique require a great deal of time and expertise to produce these datasets? This is an important question with respect to practicality, when software like Caliper takes only a few minutes to produce quantitative reports and parametric maps of lung textures on patients. 
• Given these CT scans were performed on patients, why weren't the results correlated in a multivariate regression model to determine how well the distributional analysis is associated with disease phenotypes (i.e. DLCO, 1 year outcomes, etc...)?
• Can you provide some detailed descriptive statistics and a t-test comparing normal lungs with diseased lungs from Figure 8? Also, could you provide a similar approach with data in Figure 9b? This would allow you to take the values computed that are associated with different diagnoses to see if the computed values differ by ILD diagnosis.
• Please provide absolute value standard deviations for DILD vs normal lungs in Figure 11, rather than a relative percentage of standard deviation.
• Lines 369-370 reads, "In conclusion, from a Medical Science perspective, the proposed method and model reflect in a quantitative and qualitative manner the underlying biological process". You have no data to support this statement other than the distributions that were computed plotted on a 3D plot look are different from normal lung sections. In order to make a statement like that, you would need to compare your values with pathohistology, some known biomarker, or at minimum correlate with a phenotype of the disease (i.e. pulmonary function tests). Please reword that statement or remove altogether.

Some minor points:

• Please define all acronyms the first time they are presented in the paper. For example, line 91&92 are UIP and NSIP without first defining them. 
• What is the spatial resolution of the CT images used in the manuscript? There is a comment in line 170 that says "Assuming actual pixel spacing is 0.74 mm", is that what the spatial resolution was? Also, please provide the CT imaging parameters that were used for the patients, if not, why? Were the imaging parameters different for each patient or for some of the patients?
• Line 204, describes the radial distance ≤4, I presume that is in mm? Please specify.

Reviewer 2 Report

A Novel Method for Lung Image Processing Using Complex Networks

 

First thing first, the paper needs a better polishing in writing. Specific comments:

1. Figure 3. the double line format is very confusing. Also in the figure caption, the authors need to explain why they used these particular Rds and what is the general finding, what is degree value, etc. 
2. line 250 "let us start by" ... this is an example of where writing needs to be polished for academic paper format. 
3. Figure 5. much better explanation need for these figures. what is the meaning of dot size, color, connection, relative position, etc?
4. lin 253 "(e)Degree distribution the normal sample network (f)Degree distribution the normal sample 253 network." these two are exactly the same and clearly, that is not a correct label. 

these comments did not exhaust the problems about the paper but rather showcase that the paper needs much more work to be in a form for review.