A Generic and Effective System Dispersion Compensation Method: Development and Validation in Visible-Light OCT

Round 1

Reviewer 1 Report

General comments –

The (well written) manuscript claims/implies to present a novel method for numerical dispersion compensation. However, the method described is what I consider to be the default Hilbert transform method and use (slight variation in processing detail) in my published work. I think* a large part of the OCT community would also consider the method standard. Cense et al’s method is the same but an improvement on this work, as it compensates for sample dispersion as well and doesn’t require a separate measurement, and only forms a fraction of their paper (implying limited perceived importance to community).

*However, off the top of my head I do not know where I would refer to for the basic reference on this topic (like the authors, I myself inferred/backtracked this standard method from the advanced techniques presented by Cense et al and Wojtkowski et al, which themselves are only a small part of the corresponding papers), so there maybe a gap in the literature, though it would more of a literature review and survey based.

Following from this, I don’t agree with the authors choice of terminology. Though the Wojtkowski method is the most well know piece of literature, and probably regarded as the “gold standard”, I wouldn’t describe as “traditional”, as I don’t think it is that widely implemented due to computational cost/complexity. I would regard the method described in this paper as the “default” Hilbert transform method.

I’m not convinced on the results presented, it is likely that differences between the methods is as much determined by the quality of implementation as fundamental difference between the methods. For example, a thorough enough iteration of the Wojtkowski method (with enough Taylor terms) should cover and find the dispersion/phase values of the Hilbert method. The set computational limits and optimisation method used is a subjective input.

As Photonics does have nominal novelty requirements (which I do not feel the manuscript has), I’ll defer to the editor to decide whether to veto the manuscript on these grounds. If it is vetoed, I would suggest (after addressing reviewers’ comments) sending it to one of the journals without novelty requirements (e.g. Scientific Reports) as it appears technically sound and it may fill a hole/serve a purpose/be of interest in the published literature. If it is not vetoed, my recommendation is major revisions for the authors to make the narrative and claims of the paper better reflect what is already done in the field.

Specific comments –

Title – To be reviewed at a later stage.

Abstract – To be reviewed at a later stage.

Introduction –

“It has been demonstrated that…” – too weakly stated. The theoretical basis that the axial PSF (i.e. resolution) is the Fourier transform (in k/frequency space) of the effective spectrum is a well established “fact”.

You can also generalise the examples a bit more, papers doing OCT beyond near-UV are starting emerge e.g. P. Wachulak, A. Bartnik, and H. Fiedorowicz, "Optical coherence tomography (OCT) with 2 nm axial resolution using a compact laser plasma soft X-ray source," Sci. Rep. 8, 8494 (2018). Generalising to the extreme, OCT could theoretically be done at any part of the EM spectrum.

“side lobes” – Usually, the first factor to consider if your axial PSF has side lobes is effective spectral shape (due to source of the axial PSF descried above). Apodisation or Digital Spectral Shaping will normally significantly supress axial-PSF side lobes. Polarisation as a cause would be considered after this.

“Traditional method is based on the Taylor series…” should be “Traditional method is based on iterative fitting of the Taylor series…”

“However, these methods are limited to certain orders of dispersion [21]”  - That is not true or what the paper says. You can fit as many orders as you like (or have the data quality to do). What the paper say’s is that only the 2^nd and 3^rd order terms are practically significant. (1^st order term is group delay and thus not a “dispersion” term to be corrected.)

2.1 –

Somewhere in the manuscript you need to briefly justify why you use a fibre optic setup here. When high axial resolution is the aim, the usual approach would be a free space system which will reduce dispersion mismatch and eliminate polarisation issues (note from practical experience – as long as dielectric mirrors are not used).

3.1 –

Figure 3 – Also add overlaid linear amplitude central PSF’s.

Better than average.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors should indicate in the abstract and conclusion what kind of OCT technique (time-domain / frequency-domain) is concerned by the proposed dispersion correction method.

A point should be removed line 79

The method cannot compensate the dispersion induced by the sample itself. This is a problem in certain applications (for retina imaging for exemple, due to the thickness of the sample); this could be mentioned in the paper. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

A method for dispersion compensation is presented. The authors explain that dispersion compensation is of particular importance in visible light OCT and explain quickly the different methods that can be applied. The results with different methods are compared.

The paper is well written and clear. I have a few suggestions that I think can make it easier to read.

- I miss a bit more explanation on how the new method is applied compared to the other ones. What are the new measurements or algorithms needed to apply this new method? Maybe a flow chart like the one in fig 2 of the other methods can be introduced or a table with requirements to compare between them.

- When comparing between methods, the four PSFs are compared (figures 3 and 4) but only the images obtained without correction and correcting with the new method are shown. I strongly suggest to show all the images. Also, an image-based metric could be used to show the better snr of the new images.

- In the figures, the data is shifted to compare between the methods (in figs 3, 7, 9 and 11 the data is shifted laterally and in  fig 10 it is shifted vertically). While this method can be effective for visualizing the data, at first glance, a reader may think that the depth is close to 1 mm or that sensivitiy is close to 300 dB. My suggestion to the authors is to present the data in an alternative way.

- The method is presented in section 3 and section 4 is titled discussion. However, a large part of section 4 is new results presented using the method demonstrated in 3. In section 4.1 a glass is introduced to introduce extra compensation and results applying the four methods are presented, section 4.2 shows a problem with small oscillations and suggest a way to correct it and section 4.3 shows the application of the method in a different OCT in a different wavelengt. I suggest to move all this data to the results section. Discussion section should be devoted to explore the limitations of the methodology, or to compare the results with other studies.








I have a few other comments:

- Line 65: "As for the additional depth-dependent dispersion induced by the tissue, it is usually negligible due to the shallow imaging depth in most of SD-OCT systems [21]."
Note that in anterior segment imaging depth is larger and a depth-dependent dispersion compensation is often applied, I would specify which tissue are the authors referring to.
(same for the sentence in lines 71-72)

- Line 111 "Compared with the conventional methods, the proposed method has demonstrated better..." I suggest to state clearly which conventional methods have been used conclude this affirmation.

- Line 142. The dispersion compensation methods used are listed. The authors call the first one "the traditional method" I suggest to find another way to design that method.

- Line 142. Add references for the reader to access a study where "the traditional method", the TSMMR or the SAMMR methods are explained.

- Line 223. I suggest to define a symmetry metric and evaluate a metric to show that the PSF provided by SAMMR is more symmetrical than the one by previous method.

- line 254. I suggest to demonstrate the performance of SAMMR by comparing with the other methods. In Figure 5, I suggest to show the four images, the original, the one corrected with the traditional method, the one corrected with TSMMR and the one corrected with SAMMR. This would allow quick visualization of the images and would emphasize the better performance of SAMMR.

- Figure 8 is repeating data from figure 4. I think this shouldn't be done and I suggest to present only once the data (i.e. to remove Fig 4) or to show only the results with glass in fig 8.

- Line 324. ". The traditional method is therefore limited to certain orders" Which orders? I guess that this is the reason why it cannot correct the distortion introduced theoretically. Can the authors comment on this?

- Figure 10. I suggest to review the axis, I'm not sure if the scale of radians in the thousands is correct. Also, all curves have been shifted upwards for visualization as is stated in the figure caption but... how much have the authors shifted the plots? To avoid seeing data with 200-300 dB, I suggest to use different subplots or to draw a line and start the yaxis again in the three subplots of Fig 10.






I write here some english typos that I've seen through the paper. I have not written all of them. I suggest to carefully review all text.
 
line 93. Instead of "these methods are failed to consider high order dispersion" I would just say "these methods fail to consider the higher order dispersion"

line 96. "methods that requiring the" should say "methods that require"

line 103 "the movement of the mirror in reference" I guess is "in the reference"

line 127 "of the laser" instead of "of laser"

line 139 "to avoid the saturation of spectrometer" should be "of the spectrometer"

I write here some english typos that I've seen through the paper. I have not written all of them. I suggest to carefully review all text.
 
line 93. Instead of "these methods are failed to consider high order dispersion" I would just say "these methods fail to consider the higher order dispersion"

line 96. "methods that requiring the" should say "methods that require"

line 103 "the movement of the mirror in reference" I guess is "in the reference"

line 127 "of the laser" instead of "of laser"

line 139 "to avoid the saturation of spectrometer" should be "of the spectrometer"

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors have not even attempted to address my general comments.

The work is not remotely novel yet still claims to be. Therefore there is no way this should be published.

N/A

Author Response

We sincerely apologize for overlooking your general comments in our previous response. This was an unintentional mistake on our part and we deeply regret it. We would greatly appreciate it if you could give us another opportunity to address your comments and concerns in a revised version of our paper. We value your time and expertise as a reviewer and we hope to receive your constructive feedbacks again. Corresponding changes have also been highlighted (using red color) in the revised manuscript.

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

My conclusion

I will give a positive recommendation to publish if the following changes are made.

1.       The (implied) claims on novelty need to be toned down further. A statement along the lines of “An anonymous reviewer claims to already use a very similar method, and is possible others have used/derived a similar method independently. However, to the authors best knowledge, there have been no previous publications explicitly disclosing/describing the method an evaluating its performance.” will probably do. That way if it turns up it is already in a textbook somewhere you are covered.

2.       Following this change in narrative, you can probably do some editing down of the manuscript to match.

3.       Address the comments below, excluding ones marked “No specific action required.”. Some may overlap with these points, just refer to where they are addressed.

I’ll mark as major revisions, because it is quite substantial changes, but at this stage I am happy for the (academic) editor to check the changes to his/her satisfaction.

Good look, and I look forward to seeing the manuscript published.

Cover letter comments

Comment 1 –

How you got there is interesting and fair enough, but the problem is with the method you have ended up with is something I (a reviewer picked at random) already do and think is fairly obvious. I also think that other groups would already apply the method (when they are not doing something more powerful that also corrects dynamically for sample dispersion) and would think is obvious. To defend the subjective term “obvious”, there are papers published that map the dispersion of samples which builds upon the underlying scientific basis and is much more difficult.

You have not had to “adapt” the technique to vis-OCT, it is the same as would be done for NIR-OCT or for any SD or SS OCT system at any part of the EM spectrum. “Application to”/”applied to” would be better verbs to focus on.

I think you have missed one reason to do a polynomial fit, compared to using raw phase values like you do, is that noise is smoothed out. Related to that, you don’t need to do it because the demonstration on real samples later on covers the main issue, for fig 3. best practice would be for the calibration data (calculation of dispersion correction) and image (displayed (corrected) A-Scans) to be taken separately (not sure if you did this).

Comment 2 – I suppose, with it acknowledging more work that has been done, it is a slight improvement. No specific action required.

Comment 3 – It’ll do. No specific action required.

Comment 4 –

The motivation given works if the method was for live sample dispersion correction, where computational cost matters. The problem is your correction is for the instrument only and only needs calculating once before use, therefore computational cost does not matter.

Figure S2 and S3, without displacing the A-Scans, need to replace figures 5 and 6 within the manuscript.

Comment 5 –

I can’t take responsibility for the editor and other reviewers, and I can’t refute their subjective opinions. However, I do have to make sure objective issues I have identified are addressed before I can give a positive recommendation. Over implication / over selling of novelty is an objective issue.

Manuscript comments

Title – Change “effective dispersion” to “effective system dispersion”

Abstract –

It can and should be condensed.

Digital dispersion methods are already also widely used.

Introduction –

4^th and 5^th paragraph – An important point you are missing is that these techniques correct for sample dispersion as well as system dispersion. The dispersion correction is calculated live during imaging, and will change if the sample dispersion changes. Whereas your technique only corrects for system dispersion and is static. You also only need to calculate the dispersion correction once (and not during imaging), whereas for these two techniques computation efficiency would be a benefit.

Results –

Figure 3 (c) – zoom in on central peak so reader can actually compare

It is a good idea to give it a good edit at this stage, after changes have been made, to ensure it is coherent overall.

Author Response

We are grateful to your recommendation and appreciate the opportunity to further improve our manuscript. We have carefully revised the manuscript and made all the necessary changes according to your suggestions. We hope that these revisions have improved the quality of our paper. Thank you once again for your valuable comments and support throughout this review process.

Author Response File: Author Response.docx