Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is an excellent review article on an important topic to the X-ray phase-contrast community, on the fundamental problem of capturing wave and particle effects for imaging simulations. This review article is timely and, as the authors suggest, a welcomed update to previous reviews on the topic as important advances have happened in the last decade. Especially appreciated are the flow-charts which summarize current attempts of combining coherent and incoherent effects, providing the reader a conceptual comparison of the different approaches.

My main suggestion is to expand the discussion on problems associated with sampling in the wave-optics section (section 2). The main limitation of wave-optics is the need for very high numerical sampling (as the authors only briefly state in line 133) which is a challenge when combining simulations of higher X-ray energies and detailed object models, e.g., relevant for simulating realistic clinical applications. This warrants a more detailed discussion in this review, including recent progress on solving this issue by, e.g., referring to the recent work below:

[1] Häggmark, Ilian, Kian Shaker, and Hans M. Hertz. "In silico phase-contrast X-ray imaging of anthropomorphic voxel-based phantoms." IEEE Transactions on Medical Imaging 40.2 (2020): 539-548.

 

[2] Häggmark, Ilian, et al. "Phase-contrast virtual chest radiography." Proceedings of the National Academy of Sciences 120.1 (2023): e2210214120.

 

My other comment is on the suggested approach using the “Wigner function”. The authors present this approach as a promising way of reconciling the incoherent and coherent effects, yet provide no references nor explanations for why they believe so, and it is only briefly mentioned in the “Introduction” and “Discussion and conclusions”, but is still included in Table 1 which I find strange given that the other approaches are thoroughly described and backed up with adequate references. I suggest expanding the material on the “Wigner function” approach with more detailed reasoning on why it seems promising by the authors, as well as referring to relevant literature (e.g., how it is used in visible optics simulations as stated by the authors). Here I would also mention what the challenges are to extend the approach from visible optics simulations to the X-ray regime (how does the shorter wavelengths affect the approach?)

 

Lastly, in Table 1 I would suggest explicitly mentioning the numerical spatial sampling requirements for each method in the benefits vs limitations (e.g., limitation for the classical approaches).

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

General

The underlying manuscript by Pietersoone et al. reviews the field of X-ray phase contrast simulations in the context of combining (standard) wave optics modeling with models that take particle effects into account. That is a very interesting and recent subject and definitely worth publishing. Overall, the paper is very well written, the authors have conducted an extensive and very satisfying literature review as well as summarized well the discussed works. However, I have three major points of critique which I hope they could address in a reworked version of the paper, and that is the main reason why I suggest their work for a Major revision. That said, I believe the additional effort I am asking for, will significantly boost the significance of the paper and the overall outlook for future readers. The other small corrections/comments that I have made below, should on the other hand be straight-forward to be implemented and are just minor points.

Major points

-    While I understand the motivation for introducing MC into the simulation approaches, I would like the comparison to (standard) wave optics approach to be discussed in more detail. In essence, a reader after reading the underlying review should know under which conditions, he/she should opt for a MC-based XPCI simulation and under which conditions the wave optics approach is enough. This discussion and analysis of this aspect are currently missing in the paper. For instance, which X-ray energies, resolutions, geometries, samples necessitate one approach or the other? Dark-field imaging is for instance mentioned as one reason, but my understanding is that this only affects situations when conducting imaging at large pixel sizes.
-    The aspect of computational efficiency is currently not covered at all and not even on a proof-of-concept level. What I mean by that is that the different chapters (3, 4.1 etc.) do not even touch the topic on computational efficiency and how the different approaches compare on these aspects. Also, it remains totally unclear which type of computing infrastructure is needed to run the simulations and on which time scales those are running (hours, days, weeks)? For instance, let’s say I have a 2k x 2k x 2k volume and I’d like to run a MC simulation, which algorithm/approach should I choose? Which of the works was realized on larger/smaller volumes, which took longer time to simulate etc.)? All these discussions I would expect to be part of this review because otherwise the Chapters 4.1, 4.2. are only summaries of the original works. I am aware that the authors might not have access to all these detailed information, but then at least an educated estimation based on the applied methods, should be possible.
-    The aspect of the Wigner distribution is mentioned in the introduction as a possible direction for future research, but the detailed discussion on why that is the case is missing or at least not well explained. Please explain why is it a promising direction? How does it compare to the others and what is “the possible way to reconcile”? Speaking about future, are any of the reported simulations/approaches compatible with modern GPU architectures? Do the authors expect MC simulations (and/or mixed approaches) to profit from the shear amount of machine learning algorithms or is this subject totally off?

Minor points:

-    Figures should be referenced in the text and in each figure’s caption. Please write where the figure comes from (if taken directly from the original publication) or indicate to which Chapter it refers to.

Abstract:
-    «Accurate simulation of XPCI» is repeated two times in two consecutive sentences. That is a bit bad style and the second sentence could just start with “It could also provide…”

Introduction:
-    Page 1, Line 31: I’d suggest to be more precise. “Higher sensitivity than traditional imaging” refers both to certain X-ray energies and resolutions.
-    Page 2, Line 35: “Standard X-ray images” should be more precise. I suppose, the authors refer to standardly used clinical X-ray images.
-    Page 2, Line 42: please add appropriate citation for GBI. The original authors typically refer to the technique as “edge illumination”.
-    Page 2, Line 52: Not only higher propagation distances are problematic to PBI, but also the fact that it requires a small X-ray source, i.e. coherent X-rays.
-    Page 2, Line 62: First, it would be nice to mention in 1-2 sentences what where the essential messages and/or missing parts in those previous reviews. Then, I’d suggest to rephrase “What we offer, with this review” to “In the present review, we demonstrate/investigate/analyze how simulations…”

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have adressed my comments satisfactorily.