A Feasibility Study on Proton Range Monitoring Using ¹³N Peak in Inhomogeneous Targets

Round 1

Reviewer 1 Report (Previous Reviewer 1)

The authors have responded sufficiently to all my comments and requests. Therefore, I expect this draft to be accepted.

Author Response

Thank you for the valuable comments by the three reviewers. We revised the manuscript according to the comments and point-by-point replies are attached.

Author Response File: Author Response.docx

Reviewer 2 Report (Previous Reviewer 2)

 

This manuscript describes a study of a method for estimating the distribution of multiple positron-emitting radionuclides (SA) from the time course of PET images. The paper discusses an application of the SA method to the results of irradiation to non-uniform targets. The results are very interesting for range estimation for proton therapy.

However, there are so many unclear and questionable points that make it difficult to recommend publication as is.

 

Specific Comments:

1.         Particles produced by nuclear reactions may have high energy. In such cases, they fly far away from where they are generated. The positron-emitting nuclei focused on in this paper have relatively long lifetimes, so they will decay not at the place where they are generated but at the place where they fly away. The authors said that they counted the amount of positron-emitting nuclei generated by the "T-product". Does this mean that they are counted where they are generated?

2.         I did not understand the significance of the shape of the ordinary slab phantom. The shape seems to be different from Ref. 8; Attanas et al. evaluated the effect of inhomogeneities along the beam axis. Therefore, they used a homogeneous phantom in the perpendicular direction. Why did the authors divide the last part into two parts: it is easy to compare PMMA and bone if evaluated on two separate phantoms.
On the other hand, if the authors focused on the distribution and interactions within the boundary region, such an analysis should be performed. In addition, the composition of water and PMMA is very different from that of human tissue, which does not help us to discuss the results from the MIRD phantom. If the authors tried to evaluate the effect of compositional differences on nuclide production, a phantom with different compositions at depth would not be appropriate. (We cannot distinguish that it is an effect of the proton energy or the composition of the target.)
The purpose of this experiment is unclear. If the authors cannot state a positive need of the study with the slab phantom, then I recommend that it be removed.

3.         Is the composition of the tumor the same as soft tissue?

4.         Did the authors irradiate a spherical target with a radius of 10 mm (20 mm diameter) with a 10 × 10 mm² field? If this is not a mistake, I would like to know the intention. I don't think the fact that the authors did not irradiate the entire tumor diminishes the value of the study. Therefore, it may not need to be corrected. However, it seems unnatural.

5.         How were the irradiation time and beam intensity set? Positron emitters also decay and decrease during the irradiation period. The amount of decrease depends on the nuclide, so the amount immediately after irradiation depends on the irradiation time and beam intensity. I believe that the time course is important in this study.

6.         About image reconstruction, did the authors simulate PET measurements? Or did it simply image the radioactivity distribution of positron-emitting radionuclides? If the latter, does that mean that inaccuracies due to PET imaging capabilities are not taken into account?

7.         The sentence from line 201 is repeated from line 210.

8.         What is "concentration"? Is it the relative number of positrons generated in a voxel? Or is it "something defined by Equation 1"? Is "j" ("term") is each assumed nuclide? What is "time-activity"? Is "○x" the convolution operation? What are "α" and "β"? How did you calculate Sv with β > 1.5 and use that Sv value to reduce background? The equation is not fully explained.

9.         After dividing the analysis into 4 regions, how did the authors calculate the distribution along the depth, or value of voxels?

10.     All figures are too small to discern the results.

11.     Line 243: It is unclear to which part of the image "notably sharp" is referring.

12.     Why is the dose in the SOBP region within the bone lower than the dose inside the PMMA?

13.     What exactly is "the effect of material interface and changes in its chemical composition on the projected range of protons"?

14.     Line 257: The graph in Fig. 4 is normalized for each nuclide, so it is impossible to read which of each species is more common in that region.

15.     Line 295. The authors discuss the difference in the distribution of 13N in the PMAA region and the bone region. However, no data are presented to show the difference.

16.     The authors present the lowest threshold of the reaction as the cause of the distribution of each nuclide. If the threshold is the only cause, why does 13N increase only near the Bragg peak? A graph of the reaction cross-section might help us understand.

17.     Lines 234-341. The subject of this paper would be to verify whether SA is also useful when protons are irradiated to inhomogeneous materials. This part explains and discusses the results that are not the subject of this study. It should be more concise. For example, why 13N increases near the Bragg peak is something that has been discussed in previous studies and is not a newly discovered fact in this study. This part obscures the main purpose of the study.

18.     Fig. 6. The 15-minute image shows that 13N is dominant in the region of the distal end. It would be more pronounced if PET were taken earlier. It is unclear how much the use of SA improves the accuracy of range (distal end) estimation.

19.     Why is a component with a half-life of 0, which does not exist, greatly valued?

20.     Line 430. Fig. 9 is normalized by the maximum value for each curve, so it is not possible to determine which location is estimated incorrectly. Also, the contribution of each nuclide cannot be compared. And why does close proximity of the magnitude of the contributions lead to wrong results?

21.     It is unclear what Fig. 10 is listed for.

 

Author Response

Thank you for your valuable comments. We revised according to your comments and point-by-point replies are attached.

Author Response File: Author Response.docx

Reviewer 3 Report (Previous Reviewer 3)

In this new version of the manuscript, the authors have coveniently changed the title which seems now more suitable to describe the work. I still have concerns indeed. Even though there is a minor improvement with respect to the previous published work, there is still some misleading mentioning to PET images that must be corrected, and furthermore it must be clearly stated by the authors what are the factors that can play a role in a practical implementation of PET-based range monitoring  with their method. These are my specific recommendations:

1) PET images: At lines 172-175, the authors stated that the generated images (in this work) do "resemble images created by PET systems". This is not correct and must be addressed in the manuscript. The authors are generating simulated maps of concentration of positron emitting radionuclides: even though the physical meaning of the obtained simulation is the same as what is being measured in PET, the latter type of image is the result of a complex and indirect process of physical measurement and algorithmic reconstruction, affected by many factors such as photon scattering, random coincidences, positron range. This is is far from being "similar" to what obtained in a MC simulation of nuclide activation. The authors must correct this point througout the manuscript, and clearly state that this feasibility study just gives indication of possible practical implementation in real PET systems. In fact, the obtained results are based on the application of the spectral analysis method on ROI's traced on the ideal distribution of concentration: it must be clear to the reader that the results in real practice are certainly affected by all the issues stated above about real PET images.

2) Regarding the applicability of this method to real PET-based range monitoring, the authors must add further discussion regarding (at least): (i) which type of PET system is better suited to catch the real advantage of the spectral analysis for 13N determination in-vivo (namely, in-beam, in-room, or offline methods, considering differences in timing and systems' sensitivity);

(ii) regarding the behaviour of the 13N determination at the interfaces (especially soft-tissue to bone) what are the requirements of a real PET system in term of spatial resolution in order that this method vould be applied). A comment on the fact that different radionuclide have different range in tissues must be added, and this will further influence the real applicability of the method. 

Author Response

Thank you for your valuable comments. We revised according to your comments and point-by-point replies are attached.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report (Previous Reviewer 2)

 

I thank the authors for their courteous response. While the authors have adequately addressed many points in the revised manuscript, some points were not adequately answered. Also, the unclear purpose of the paper leaves some points difficult to comment on for proper revision: if SA is to be adapted to heterogeneous materials and shown to be calculated correctly, a quantitative evaluation and discussion of errors is needed (also, the title may need to be revised): if the goal is to show that SA improves the performance of range monitoring in heterogeneous materials, then proof of this is needed. However, neither of these is stated in this manuscript, nor in the introduction section.

Thus, it is difficult to recommend that the manuscript is published in its entirety, even at this time.

 

Specific comments,

1.         If the energy of the generated positron emitters is low and there is little difference between where it is generated and where it decays, it shall be better to state that in the manuscript.
It is difficult to understand that they are counted in both. Could the author please explain it in detail?

2.         I had overlooked the use of multiple phantoms for the paper by Attanasi et al. I would like to thank and apologize to the authors.
However, even at this time it is unclear why this phantom was employed. The authors claim to be discussing boundary effects, but I could not figure out which sentence applies to them. The “boundary effects” are, for example, the sharpness of the boundary region and the interaction between the PMMA and Bone region in the distal side of the Bragg peak in the bone region that would not be seen alone. I was not simply referring to discussing the difference between bone and PMMA.
If the authors simply discuss the difference between the two materials, the authors should calculate on separate phantoms. Even if a study was conducted on a single phantom, as in this case, it is possible to create 1D profiles separately for the PMMA region and the bone region. Neither would be difficult. Why did the authors make just one 1D profile for all regions and discussing material differences?

3.         “Since we are dealing with slab phantoms......." in the response #2. I did not understand what this sentence was trying to assert. I understood that the authors used the same phantoms as claimed by Attanasi et al. However, it is unclear what their experimental results reveal with respect to clinical applications. There are many explanations given in the manuscript about the distribution of the produced RIs. And this explanation is probably correct. But this is not what the study reveals. I think we know this from previous studies.
Also, if one wants to discuss the distribution of proton energies and produced nuclides, heterogeneous targets are not appropriate.
The effect of heterogeneity on the distribution of positron emitters is a very interesting topic. It may be necessary to conduct research on a simple phantom before considering complex systems. But what does this slab phantom substitute for human complexity? If the authors wants to consider the influence of target material and reaction threshold on the distribution of produced nuclides (as in the discussion of line 266~), they should use homogeneous targets in each material. Currently, only a mixture of the two effects can be observed. There is no discussion of the effects and interactions of inhomogeneities at the boundaries, either in the direction of the beam or in the lateral direction.
In fact, the results of this phantom experiment have not been used in the discussion to better understand the results of the calculations with the MIRD phantom. The distribution of 11C near the distal end of the proton beam is completely different between the slab phantom and the MIRD phantom. This could indicate that slab phantoms (especially PMMA) are not suitable for simulating the heterogeneity of the human body. Once again, I would recommend that the paper be written only for the MIRD phantom study. The section is very long because it describes the results of this slab phantom experiment, making it difficult to understand what the paper is trying to claim.

4.         I did not understand what the response #5 was trying to assert. I am not asking about the number of the incident protons in the MC calculation. I wanted to know what kind of irradiation is assumed. The time effect can be calculated with the optional function of PHITS. Or it can be easily calculated by exponential function instead of MC calculation.
In any case, perhaps the response means that the authors did not consider the decay of radionuclides during treatment irradiation. If so, I think that should be stated in the text and its impact should be added to the discussion.

5.         The answer #9 was inadequate for me. Of course, the distribution of nuclides (the value of voxel) would be calculated taking into account the results of the calculations for each region (divided four regions). I would like a careful explanation of the methodology how do the authors take the SA results into account in the distribution calculations? In other words, I would like the author to explain how "Using these data, a single voxel image for each ROI was generated (p. 6, LL. 230-231)".

6.         About response #12. Normalization does not result in a lower dose. Also, even if the fluence of the protons with maximum energy is lower than that of mono-energy proton, it is the same for PMMA. I don't think this answer is an explanation. Is the dose to the bone region in the target also lower in general treatment for some reason?

7.         Response #14. For example, in lines 275-277 the authors state "the amount of ¹⁵O radionuclides would be higher compared to ¹¹C radionuclides in the water filled region." Other parts of the paper also discuss the comparison of the amounts of radionuclides. If the authors are comparing quantities, then the graph of normalized values is inappropriate.

8.         Response #15. Figure 4 only shows the total distribution of the two regions. While I could see the authors' point, I don't think it is "clearly presented in Fig. 4" from a scientific point of view. Why not separate the two graphs?

9.         The authors claim that the "Important objective is the formation of a 13N peak in an inhomogeneous slab phantom" in Response #17, but the material of the slab phantom is almost uniform near the Bragg peak. The effect of the boundary between bone and PMMA on the RI generations is also not discussed at all. The authors claim that the generated positron-emitters do not travel far. If this is true, then the generated amount of each nuclide can be calculated simply by the composition at a given location and the energy of the protons at that position. Non-uniformity along the beam axis, from a macroscopic point of view, (and more specifically, the composition of the material upstream of the beam) is not a problem in the distribution of positron-emitting nuclides. It is unlikely that any new discoveries were made.
Also, even if this subject were important for this manuscript, the discussion in lines 242-317 is making a directly irrelevant argument. I would suggest that the text be substantially revised to clarify what is new.

10.     Response #18. I do not understand the purpose of this study. The authors state that "The improvement of accuracy of range estimation is not the task of the SA method shown in this paper.” Then what is the purpose of SA? It is difficult to say for sure because all the graphs are normalized by nuclide, but if the production of 11C and 13N is comparable, it should be possible to estimate the range using only PET images at 10 to 15 minutes (without SA). For example, in Fig. 5(b), the proton range is very clear. The authors state in lines 410-412, "These results show that ..... using our proposed SA technique.” However, I believe that the 15-55 minutes of images show that the range can be estimated WITHOUT SA. What contribution does SA (or extract the distribution of 13N) make to range monitoring?

11.     Regarding response #19. The authors state that they performed SA on distribution of positron emitters calculated by MS. What is the “background” under the conditions? Are there that many radionuclides with half-lives long enough to be considered infinite in this study? If so, it would be better to give specific radionuclides.
I don't think it matters where the radioactive isotopes are counted, it does not affect the SA calculations.
The authors say in response #1 that the energy of produced RI is low and does not travel far, but this comment states that the migration of produced RI has an effect. (I never understood what they meant by "Larger regions that generated radionuclide cannot reach in the region of interest" to begin with.)

12.     Regarding response #20. The authors state that the SA and MC do not coincide near the entrance. However, these are both normalized curve comparisons. For example, looking at the 11C distributions in Fig. 9(a) and (c), we can see that the normalizing points are different for MC and SA. A correct comparison would be difficult in this situation. In fact, we cannot rule out the possibility that the estimates near the entrance are in agreement and the estimates near the normalizing point are very different.
As a general, I think it is easier to be led to wrong results by errors in those with smaller contributions than in those with larger contributions.
We need evidence to support these authors' claims.

13.     It has been shown in previous studies that 3-dimensional plots are possible (it means that there is no novelty to show the 3D images). In this study (although the purpose of the study is not clear) there is no contribution to the study from this 3D plot. I recommend that it be removed.

Author Response

Thank you for your valuable comments. We revised our manuscript according to your comments and suggestions.

Author Response File: Author Response.pdf

Reviewer 3 Report (Previous Reviewer 3)

Even though I still judge this paper of scarce real-world applicability due to the several neglected physical factors (especially regarding PET imaging), the authors have amended the manuscript as requested. I have no other requests.

Author Response

Thank you for reviewing our manuscript. 

Author Response File: Author Response.pdf

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

This paper is the first application of SA approach to an inhomogeneous target. This approach is almost the same as the authors previous paper “M. Rafiqul Islam et al., PLoS ONE 17(2): e0263521”. Therefore, in order to publish this paper, some additional discussions, which could be made because of the use of inhomogeneous targets, should be added.

 

I think the following revision should also be made:

 

In Fig. 3, differences between the pristine Bragg peaks and the peak positions of the N-13 seems different between the PMMA and the bone. The difference for the bone seems shorter than that for PMMA. Is this reasonable? Please add a discussion about this issue.

 

The followings are additional revisions:

 

Line 86: Perhaps, “direSction” is typo. It should be revised to “direction”.

 

Line 92: “between the proton beam” should be replaced to “between the source of protons”.

Reviewer 2 Report

 

This paper describes a method for estimating the amount of each nuclide emitting the PET signal (annihilation γ-rays) by using Spectral Imaging (SA) for the time-course of the PET signal to confirm the range of clinical proton beam measuring on an inhomogeneous target.

The measurement of a particle-beam range is an important issue in particle radiotherapy, and PET imaging for the issue is expected to be of great interest to readers.

However, the purpose of the study is unclear in this paper. It is not clear for what purpose the study was conducted on inhomogeneous targets or what issues were resolved from this study for clinical application. If clinical applications are to be considered, the phantom should be an appropriate soft tissue composition rather than water or acrylic. Details of the experimental methods are also unclear. Therefore, it is difficult to recommend publishing the manuscript as is. Specific comments are as follows;

 

General

1.         There are many typographical errors and inappropriate use of abbreviations (especially SA). It needs to be proofread carefully.

2.         The structure of this article is not well organized.

3.         The amount of positron nuclides produced depends on the composition of the target. However, in this study, their distribution is very different from that of the human body. Nitrogen is particularly important in this paper, but it is only present in some regions.

 

Introduction

4.         The Introduction is relatively long but does not show the characteristics of SA or the benefits of applying it. It also does not describe the problems of previous studies and the purpose of using inhomogeneous targets in this study.

5.         Will the reaction ¹⁶O(p,2p2n)¹³N occur instead of ¹⁶O(p,α)¹³N?

 

Material and Methods

6.         Line 94: The number of particles used in the calculation (10⁹) cannot necessarily represent the accuracy of the calculation. For example, is weight set? How is the calculation grid distributed? (If there are other settings that are important for the calculation, they should be indicated.)　It could be a good idea to show how much error there is in the calculation results.

7.         Why were the lung and bone compositions set the way they were? If you have references, it would be better to show them.

8.         Why did the authors choose this phantom arrangement? It does not look like it is simulating the lung region of the human body. If the lungs are to be simulated, why did the authors divide the last region into two parts? It would seem that “soft tissue including bone -> lung -> soft tissue including bone” would be a more appropriate arrangement to mimic a human body.

9.         Lines 94 to 98 could be written in the Introduction section.

10.     How were the irradiation time and dose? These are necessary information to make an assumption about the PET image to be obtained.

11.     Fig. 1 (left); the axes are distorted.

12.     It would be easier to compare the result figures with Fig. 1 (right) if the units in Fig. 1 (right) were unified in mm. It would also be easier to understand if you write the x/z axes on the right side.

13.     How were the PET images created?

14.     Line 100: Does “t-yield” count the positron nuclide generation at the location where it occurred? Or is it counted at the position where the radiated nuclides stopped?

15.     The authors do not describe what kind of analysis was done with “PyBLD” or “AMIDE”.

16.     Lines 115-26 should be described in the introduction section.

17.     Line 137: Does “concentration” mean radioactivity concentration (Bq/g, etc.)? I don't understand why this is calculated by the convolution of the time-activity (radioactivity at that time?) and the time-decay term.

18.     What is the “term” in line 139?

19.     Line 139 says “generally between 100 and 1000”. How many values were actually used in this study?

20.     What range and interval did you analyze for “β”?

21.     What is "frame" in line 155? And what does it mean to adapt to 40 frames? In the next sentence, it says that it was calculated for 55 frames. Why did the authors change the number of “frames”?

22.     Lines 160-161: I do not understand the meaning of "transformed into", "scheme", and "obtained results". I do not understand what they meant. Perhaps that is why I could not understand the specific estimation method in this study.

 

Results and discussion

23.     Line 177: What does "beam hardening" mean for proton beams?

24.     Fig. 3: How did the authors normalize the data? Is the maximum of somewhere used?

25.     Fig. 5: In all regions, there seems to be a very large component of “half-life 0”. Could this mean that this method could not accurately estimate the distribution of each nuclide?

26.     Considering the shape of the phantom, I think the change in the y-direction is small. Why did the authors plot them in 3D? I find the distribution confusing.

27.     Line 237: "close to" is not a quantitative expression.

28.     Fig. 8: Why are the abundances of all nuclides less in the shallow region? Would the results accurately reproduce the number of positron emitters?

29.     Line 251: The authors state "lead to the proton range verification". If we only want to know the range, according to Fig. 3 and 7, the nuclides other than N-13 that can be made upstream of the range are not important and can be estimated from the total positron nuclide distribution, can they not? Please clarify the significance of using this method (SA).

Reviewer 3 Report

This work from Islam et al presents a method for proton range verification in hadron therapy from Positron Emission Tomography data, based on spectral analysis. All results presented are based on Monte Carlo simulations. The final claim is that the spetral analysis approach can be successfully applied to separate the signal of 13N from those of 15O and 11C in PET images, which is useful for the determination of proton range.

The authors have recently published a very similar work, where the spectral analysis approach has been tested on a homogeneous phantom (Islam M.R. et al, Appl. Sci. 2022, 12, 823. https://doi.org/10.3390/app12020823). In that work, experimental data from a prototype Positron EMission Mammography (PEM) system were presented along with Monte Carlo simulation. Honestly, I don't believe there is sufficient new information in this presented manuscript to justify another publication. Moreover, the claim that the same method could be applied to inhomogeneous samples would require at least a full simulation (or, better, an experimental measurement) of a real PET acquisition: instead, the authors are using the result of activation from proton beam coming from the MC simulation as if they were equivalent to PET activity data. This is far from being true. For instance, medium attenuation of 511 keV photons is totally disregarded in this work, which is the main concern when working with inhomogeneous samples. Furthermore, there is no discussion at all on how the spectral analysis approach could handle the biological washout in real living subjects. For these reasons, I could not recommend the publication of this paper.