HTN: Hybrid Transformer Network for Curvature of Cervical Spine Estimation
Round 1
Reviewer 1 Report
The authors are tried to attempt article entitled" HTN: Hybrid Transformer Network for Curvature of Cervical 2 Spine Estimation".
: Due to long-term desk work or unhealthy lifestyles, an increasing number of young peo- 9 ple have suffered from cervical spondylosis in recent years. Early diagnosis is crucial for curing 10 cervical spondylosis. The Cobb angle method is the most common method for assessing spinal cur- 11 vature; however, manually measuring the Cobb angle is time-consuming and heavily dependent on 12 personal experience. To avoid the risk of human error, it is important to establish an effective, aux- 13 iliary diagnosis model. In this paper, we propose a fully automatic system for the measurement of 14 cervical spinal curvature on X-rays using the Cobb angle method, which can reduce the workload 15 of clinicians and provide a reliable basis for surgery. A new network, the HTN, which blends a self- 16 attention mechanism, self-supervision learning and feature fusion, is proposed. To evaluate our 17 method, we proposed a cervical spondylosis dataset referred to as CS799. Our model can achieve 18 an SMAPE of 11.06% and a significant Pearson correlation coefficient of 0.9619 (p<0.001) on CS799, 19 which demonstrates that the results of our model have a strong correlation with those of the ground 20 truth annotation. The absolute difference between the ground truth and the prediction obtained is 21 less than 2°, implying clinical value. Statistical analysis proves the reliability of our method for Cobb 22 angle estimation. To further prove the validity of our method, the HTN is also trained and evaluated 23 on the public AASCE MICCAI 2019 challenge dataset. The experimental results show that our 24 method can achieve a comparable performance to that of state-of-the-art methods, which means that 25 our method can measure not only the curvature of the neck but also the curvature of the entire spine.
1. Authors need to update by adding recent literature.
2. Why the author considered the hybrid model and state of art must be provided in the introduction.
It can accepted after answering above queries. Minor revision.
Author Response
Attached is our response letter.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript has some major limitations regarding its structure and its length. First of all it should be clear and well declared the type of manuscript (review, communication, original article,…) in order to understand the main purpose of the authors. Since it seems an original article, the paper should be shortened of at least 800 words, being more adherent to the title and reducing the citations in discussion. The introduction is definitely too long, containing some parts, especially after the figure 1, which should be removed or recontextualized in the discussion section. A clear definition of “transformer” and what characterize a hybrid one is missing, but it is necessary since most of the readers are not accustomed to this topic. In methods section the paragraphs 3.3 and 3.4 present results and all the subsequent text and figures should be restructure in a Results section, which is missing. Please remove and replace in an appropriate results section. The paragraph 3.4 should be renamed since “compared..” is inappropriate. Overall, due to this structural and some linguistic mistakes, the paper is difficult to be read, appearing chaotic. Without a deep revision, with an extended and well-structured enunciation of results, a more organized distribution of paragraphs and definitions, and a careful revision by a native English speaker, I wouldn’t consider this manuscript for publication.
Author Response
Attached is our response letter.
Comment 1: The manuscript has some major limitations regarding its structure and its length. First of all it should be clear and well declared the type of manuscript (review, communication, original article,…) in order to understand the main purpose of the authors. Since it seems an original article, the paper should be shortened of at least 800 words, being more adherent to the title and reducing the citations in discussion.
Response 1: We have carefully revised our manuscript according to your comments. As you mentioned, this is an original article. We have modified the abstract of the manuscript. (Please refer to Line9-Line24). Moreover, we think keeping some relevant works in the introduction is necessary. We tried to remove redundant paragraphs and reduce unnecessary citations to clarify the article structure.
Comment 2: The introduction is definitely too long, containing some parts, especially after the figure 1, which should be removed or recontextualized in the discussion section.
Response 2: We have removed some redundant paragraphs and wordy sentences in the introduction to make the article more reasonable and logical. Besides, we have rearranged the structure of the manuscript.
Comment 3: A clear definition of “transformer” and what characterize a hybrid one is missing, but it is necessary since most of the readers are not accustomed to this topic.
Response 3: We added references [3-6] about transformers and stated the characteristic of transformers in the introduction (please refer to Line82 -Line106). In addition, we explained this in detail in Section 2.3. We think Section 2.3 is sufficient to explain the feature extraction and computation process of the transformer. Besides, we wonder which hybrid structure would lead to better performance. Therefore, we explored the impact of introducing a transformer earlier in the encoder-decoder network on the final model. In other words, we removed residual module 1 in the encoder-decoder, which resulted in a larger size of feature maps processed by the transformer. (Please refer to Line193-Line196, and Figure 2).
Comment 4: In methods section the paragraphs 3.3 and 3.4 present results and all the subsequent text and figures should be restructure in a Results section, which is missing. Please remove and replace in an appropriate results section.
Response 4: Thank you for your suggestion. We merged the contents of Appendix A into Section 3.3. (Please refer to Line355-Line367). We also moved Appendix B to Section 3.6. (Please refer to Line416-Line431). In addition, we bolded the best results and emphasized them using a green background in Table 1 and Table 3.
Comment 5: The paragraph 3.4 should be renamed since “compared..” is inappropriate.
Response 5: We modified the subheadings of Section 3.4. (Please refer to Line368).
Author Response File: Author Response.pdf
Reviewer 3 Report
I found the article "HTN: Hybrid Transformer Network for Curvature of Cervical Spine Estimation" of a good overall quality and I have only some minor comments:
lines 124-125: "[42, 43] uses transformers to extract global contexts in the field of medical image segmentation." - I don't think that is appropriate to start sentence direclty with citation
line 127: "Inspired by the idea proposed in [42, 43]. We fuse a transformer module..." - This probaly should be one sentence
line 500: "...ensure the reliability of computed aided diagnosis results[48]。The impact of lighting..." - In the pdf file that I got, after citation, wrong type of dot is used
line 518: "...should pay more attention to it. 43 Although our experiments..." - The "43" here is a citation?
Author Response
Attached is our response letter.
Comment 1: lines 124-125: "[42, 43] uses transformers to extract global contexts in the field of medical image segmentation." - I don't think that is appropriate to start sentence direclty with citation.
Response 1: Thank you for your suggestion. We have revised this sentence as follows. (Please refer to Line96-Line97).
'Chen J et al. [29] and Wang W et al. [30] use transformers to extract global contexts in medical image segmentation.'
Comment 2: Line127: "Inspired by the idea proposed in [42, 43]. We fuse a transformer module..." - This probably should be one sentence.
Response 2: Thanks for pointing out our mistake. We apologize that we used the wrong punctuation. We fixed punctuation errors and other grammatical and spelling errors. (Please refer to Line99-Line100).
Comment 3: Line500: "...ensure the reliability of computed aided diagnosis results [48]. The impact of lighting..." - In the pdf file that I got, after citation, wrong type of dot is used.
Response 3: We apologize that we used the wrong punctuation. We fixed punctuation errors and other grammatical and spelling errors. (Please refer to Line445-Line447).
Comment 4: Line518: "...should pay more attention to it. 43 Although our experiments..." - The "43" here is a citation?
Response 4: Actually, "43" was typed by us by mistake. We have already modified this error. (Please refer to Line464-Line465).
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The manuscript still lacks of sufficient criteria making it suitable for publication
Author Response
Dear Reviewer,
Submission of the revised version of our manuscript entitled "HTN: Hybrid Transformer Network for Curvature of Cervical Spine Estimation." (Manuscript ID: applsci-1927808)
We highly appreciate your time and consideration. We have carefully considered your comments and have made the following revisions to the manuscript. Changes in the manuscript are shown highlighted in red color in MS Word.
- We have tried to remove redundant paragraphs and reduce unnecessary citations to clarify the article structure. In addition, we added some conjunctions and pronouns to make the article more logical. Finally, we also have rearranged the structure of the manuscript again. (Please refer to highlighted in red color in MS Word)
- We have rewritten the contribution part in Sec.1 and the conclusion. To facilitate the development of this field, we will release our code public and pre-trained model. However, due to data privacy, our datasets cannot be released. (Please refer to Line 90-Line 114, Line 447-Line 450)
- We have double-checked minor errors and fixed them to make the manuscript suitable for publication. Minor errors mainly include spelling, grammar, and comparatives.
- For objectivity, we've reduced the use of first-person pronouns as much as possible in the updated version by using passive sentences and subject elision. Some first-person pronouns cannot be modified. Otherwise, it will affect the readability of the article.
With this revision, our manuscript has been reduced by two pages (from 16 pages to 14 pages). For the completeness of the method, we cannot reduce the description Sec.2. If you have other concerns, please let us know. Thank you for being so patient.
Best regards,
<Yifan Yao> et al.
