Application of the Approximate Bayesian Computation Algorithm to Gamma-Ray Spectroscopy

Reviewer 1

The manuscript describes further developments of the approximate Bayesian computation (ABC) algorithm for

peak location and net area computation, as well as its potential application for the radioisotope identification

(RIID).

The authors’ contribution is novel, it generally looks technically sound and of practical importance.

The authors provided a good overview of previous works in the subject area. The manuscript is illustrated

by numerous graphs and numerical examples. The conclusions largely sound reasonable and justified.

In the same time, most of the descriptions appear too general, superficial, often redundant and rather vague.

On its clarity, consistency and style, the manuscript cannot be considered as adhering to the highest

standards of scientific publications. It would be desirable to revise the manuscript in order to make it more

comprehensible and more useful for potential readers, especially for those who will probably like to reproduce

the described algorithms and results. I anticipate that readers may have difficulties to fully comprehend

and appreciate the author’s original work described in Section 6. A more detailed and clear description

of the suggested new features in the ABC algorithm and its application to the RIID would therefore be desirable.

>Done. We expanded sections 5 and 6 as suggested.

Another general observation is that the manuscript does not seem to be either a pure review or a

pure research paper.

>Noted. We opted to expand the new material in section 5 and 6 and so the revised paper is a research paper. The review

sections are shortened and we note also that reviewer 3 does not like the original style of our review. We assume readers

are familiar with the previous review papers and so the “review” is revised to be background and context, with a very brief

Discussion of progress on RIIDs with NaI detectors since the 2009 [7] or 2019 [11] review.

It would be beneficial that the authors decided on and stick to either of these options.

> Good suggestion. It is now a research paper so the “review” is revised to be background and context, with a very brief

discussion of progress on RIIDs with NaI detectors since the 2009 [7] review.

,

In my opinion, it would be logical that the original part presented in Section 6 becomes more elaborated

and detailed; while the material presented in Sections 3, 4 and 5 be significantly reduced or even omitted.

>Excellent suggestion. Done. We reduced sections 3-4. However, there is new material in Sections 5-6, so the

paper is a research paper.

Further specific comments and remarks are given below.

1.     Title: The title looks too generic and not entirely consistent with the content of the manuscript.

2.     It would be desirable that the title reflected the main novelty presented in the manuscript

3.     (as far as I understood it), which is the demonstrated application of the ABC based algorithms to RIID.

>Thanks for this comment. The revision makes it clear that the paper’s focus is peak location and

>estimation as a key step in some RIIDs, but the extended section 6 also has new RIID results, so the revised title is:

>Application of the Approximate Bayesian Computation Algorithm to Gamma-ray Spectroscopy

4.     Keywords: “Peak detection” would be better to replace by the “peak location”

5.     and “peak area determination”.

6.     > Agreed. Done

7.      

8.     Throughout the manuscript: The choice of Ac-225 as an example looks rather artificial

9.     and impractical. It would have been better to make a more useful (from the point of

10.  view of a typical RIID use case) selection of a radionuclide, e.g. Ra-226, Th-232, uranium,

11.  or any of the commonly used industrial and/or medical radioisotopes.

>Noted. Changed. Now using Lu177m versus WgPu as first RIID example and opted to use I-131 for Figures 1&2. Revision

>mentions that I-131 is a medical isotope that is sometimes detected at US borders. Figures 1 and 2 are simply examples

> with real and simulated data.

12.  Throughout the manuscript: Section 5 seems to be mistakenly mentioned instead of Section 6. Please check.

>Agreed. Done.

13.  Throughout the manuscript: It is advisable to use the term “full energy peak” instead of the “photopeak”.

>Done.

14.  Sections 3 and 5 barely show any data related to the application of the ABC algorithm to the

15.  peak location and area estimation (as one would expect from the information stated in lines 18-20

16.  of the Abstract). If it was the intention to demonstrate the capability of the ABC approach for

17.  peak location and area estimation, respective testing results are to be presented.

>More testing results and spectra have been added.

18.  The use of modelled spectra: Please clearly describe the approach used for spectrum modeling.

>The revision clarifies that we only report on GADRAS modeling without and with model bias added

19.  Provide a description of the need addressed and the purpose for which the modelled

20.  spectra were utilized in the study.

>Done. Added to Summary section.

21.  Line 27: It is not clear what “high dimensional gamma spectra” mean.

22.  >Clarified. The context is that ABC can use much lower dimensions than the 1024 (or other large number) of channels.

23.   

24.  Line 49: Change “neutral” to “neural”.

25.  Lines 111 – 135: The usefulness of the provided example of the code is questionable. More comments

26.  may be needed. Also, please indicate the programming language used.

27.  > Agreed. Dropped the R code and used f(x) notation. Still cite R [29] because all results were obtained

28.  from programming in R.

29.  Page 7: Figure 6 does not seem to be mentioned in the text. Please check if this figure is actually required.

30.  > It was, but the location of Figure 6 was confusing. Now the reference is closer to the Fig. 6.

31.   

32.  Before its description in Section 6, the ABC algorithm is mentioned many times in the preceding sections of

33.  the manuscript. This is rather confusing.

>Noted. We moved the ABC description to new subsection 5.1 just before its first usage.

section 6 because we tell readers that ABC description is in Section 6 while “selling” its use prior to that.

34.  Line 319: Change “0^4” to “10^4”.

35.  > Thanks.

36.  Line 330: Please explain what is meant by the “broader background”.

37.  > Thanks. We edited to the background as modeled by P₁-P₄.

38.  Figure 7: Please indicate the true peak location in the given example. Which of the approaches

39.  (i.e. with or without smoothing) gives a more accurate estimate for the peak location

40.  > Thanks. Done. Expanded this example and in this case, smoothing leads to smaller variance but raw and

41.  >smooth lead to about the same accuracy.

Comments and Suggestions for Authors

The manuscript was very difficult to read. Outside of the relevance and scientific quality of the work described, in

too many instances the means by which the work was reported affect too much the ability of the reader to clearly

understand what is being described/proposed. I would therefore advise the authors, in the spirit of their intension

to provide a review paper, to clearly take the reader from A to B, where A is the collective work that has been done

so far of the subject, introducing and commenting much more on the relevance and short-coming/limitation,

accomplishments of the past work, to B, which is their proposed approach.

How is B relevant and of value to gamma spectroscopist compared to other approaches ?   

The work describes revolves around a couple of examples (Ac-225/Pu) with no introduction as to why

these were chosen, how relevant they are (to any specific application??), how the proposed work compare with

other means of performing RIID...

>Because of reviewer 1 and 3, the revision is a research paper. There is still discussion (very brief) of other RIIDs, but the 2

>previous reviews [7,11] are assumed to be available to the reader. Also, changed Fig 1 and Fig. 2 from Ac-225 to I-131

>per this comment and per reviewer 1 request to use a more common medical isotope such as I131. Note also that other nuclides are added in

> the expanded section 6.

The numbering in the manuscript is incorrect. References to ABC are made throughout the manuscript where the

ABC is only described in Section 6. In general the overall structure of the document is to be reconsidered. 

>We fixed the typos regarding section 6. We moved the ABC description to new subsection 5.1 just before its first usage.

Maybe it is just me, but I do not like to read '[X] did that'. I much prefer reading 'In their work, XXXXX et.al did that [X]'

>If paper accepted, final proof stage makes choices such as that. No strong preference among authors,but lead author

>prefers minimalist acceptable sytle.

Overall, the work presented in the manuscript is undoubtedly original work and has some merit in being reported.

> We agree and because of reviewer 1 and 3 comments, we changed to Research Article and the revision makes it clear

> that there is new material in sections 5-6 and a short linkage to the literature on other non-ABC options in use.

However, the way the manuscript is reporting it is too often badly constructed/disjoint and even incoherent. This

leaves the reader the very hard task of having to make sense of statements which often appear as unjustified/random.

I would therefore strongly recommend for the manuscript to be returned to the authors for them to provide

major corrections to the manuscript.

> We think the revision is coherent. The revision describes ABC in new section 5.1 before its first usage.

I could try to provide some further examples of where such corrections would be needed, if you think that it

would be beneficial. Let me just mention simply here a few things, like the use of references in the text without

mentioning the people involved yet claiming to be a review paper and providing very little insight into the value

or shortfall of the work previously done - only that it has been done.

>The review aspects are downplayed. But we think the intended readers will know about [7] and/or [11]. It is a Research article.

>Your comments are very helpful.

The numbering of paragraphs which does not match the numbers used in the text on many occasions.

>Edited as many as we caught. If accepted, this is fixed during proof stage.

Whole paragraphs dealing with introduction material, not placed in the introduction, overall confused structure of the paper.

No clear broader definition of the application (RIID) of the discussed analytical technique/algorithms with

ad-hoc examples taken from both low (Ac-225) and high resolution spectra (plutonium) without

commentating on why these examples were used, their relevance/differences and overall how the

described algorithms would be better (or worse) than what is currently done...

>The revision makes it clear that the use of ABC is novel, partly because it allows for relatively easy experimentation regarding

>what features (summary statistics for ABC) from the spectra are used (all energy bins, some energy bins, peak locations,

>peak areas etc.). Both low and high resolution can use peak areas, so the peak area estimation and location is in section 5; we do

>not consider ultra-low resolution spectra such as plastic scintillators that rarely exhibit photopeaks.

> The RIID topic in section 6 is only for low resolution because that remains a priority in fielded instruments.

>The revision makes it clear that reader is directed to references that describe

>efforts to improve RIIDs because it is believed that NaI detectors will continue to be fielded and that improved algorithms are

> worth the effort.   Also the revision mentions that Ac-225 is sometimes detected as a medical isotope at border crossings, so is

> retained in Fig. 1 and 2, but OMITTED in the ABC-based RIID section 6 where more challenging and interesting cases are presented.

Lack of adequately defining in some mathematical expressions what the parameters are,

> Opted to drop the R function and just use f(x) in Eq. (1).

and too often to many assumptions

that the work reported elsewhere is known of the reader, to make any of the comments made meaningful

(that is not just that anybody should necessarily be able to understand what is written, but that even people

with an appropriate background will struggle to make sense of what is written for the text uses the same

notation/assumptions that are used elsewhere without stating clearly what they are)...

>We agree that the intended readers are those who have read or can read [7] and/or [11]. That way the paper is not so long.

>This is another taste issue.   The main points for intended readers are (as revision now states   near new line 45):

>there is reason to believe that RIID

>performance using NaI detectors (most common in field) can be improved. ABC is shown to perform well (examples expanded

>in revision) in peak location and area estimation and in the initial RIIDs considered (expanded examples).

> Also, the abstract is revised to make all this as clear as possible.

>Thanks for taking the time to give constructive suggestions. We do intend the readers to have at least the [7] and/or [11] references

>available in order to keep the paper at about 20 pages