Brain Extraction Methods in Neonatal Brain MRI and Their Effects on Intracranial Volumes

Round 1

Reviewer 1 Report

The authors assess the efficiency of different brain extraction methods in evaluation of intracranial volumes. The manuscript is well-written and detailed. However, my concern is that although the authors have done extensive statistical analysis to make conclusions, their sample size is very limited (only 22). Hence, they should be careful when inferring efficiencies of different BE methods with such a limited sample size. 

Could the authors please explain why they did not choose iBEAT2 or U-NET as the deep-learning methods for BE? 

The title should be edited to make it more clear and crisp.

Please correct errors on lines 221 and 240. 

Author Response

Dear Reviewer,

thank you so much for your attention to our work. 

please find in the attachment our responses to your comments.

Kind regards

Author Response File: Author Response.pdf

Reviewer 2 Report

The study introduces and evaluates different BE methods in MRI scans of preterm neonates, specifically focusing on their impact on ICV estimation. While manual segmentation is considered the gold-standard, this study provides valuable insights into the performance and accuracy of automated methods in this specific population.

1. Consider expanding the introduction to provide a brief overview of the challenges associated with neonatal brain MRI beyond brain extraction, such as motion artifacts or image resolution issues. This could help readers appreciate the broader context.

2. The study's sample size of twenty-two premature neonates is relatively small. While the results are promising, a larger sample size would enhance the generalizability and statistical power of the findings. Discuss the potential impact of the small sample size on the reliability and representativeness of the study's conclusions.

3. The paper acknowledges the presence of inter and intra-rater variability in manual segmentation, which is considered the "gold-standard." However, the study does not provide details on how the variability was addressed or controlled for. Elaborate on the steps taken to address inter- and intra-rater variability in manual segmentation. Discuss the potential impact of this variability on the reference brain masks used for evaluation and the potential implications for the comparison of automated methods.

4. The study does not extensively compare its results with those of previous studies that evaluated BE methods in neonatal brain MRI. Providing a more comprehensive discussion on how the findings align or differ from previous research would enhance the contextualization and interpretation of the study's outcomes.

5. Acknowledge the absence of neonates with detectable brain lesions or congenital conditions in the study cohort as a limitation. Discuss the implications of this limitation on the generalizability of the findings to a more diverse clinical population.

6. Explicitly mention the exclusion of neonates with significant motion artifacts or poor image quality. Discuss the potential bias introduced by excluding such cases and how the findings might differ in real-world clinical scenarios with varied image quality.

Author Response

Dear Reviewer,

thank you so much for your attention to our work. 

please find in the attachment our responses to your comments.

Kind regards

Author Response File: Author Response.pdf

Reviewer 3 Report

This work compares different MRI brain extraction methods on neonates. A total of four methods are compared with respect to manual segmentation.

Overall, the paper is well-written and clear. The methodology is sound, well-justified, and described in detail. The results are interesting and useful. The discussion covers all the relevant points observed across the paper.

For these reasons, I suggest to accept this paper after the following issues are addressed. 

* Major issues

- Expand the description of the regression analysis performed in section 3.3.2 (line 258). In particular, clarify which one is the dependent variable and which one are the independents variables.

- I find Figure 4 very interesting and useful to understand why and where the methods produce different results. From what I can observe, the methods produce very similar results in the first two slices and different results in the third slice. Please add a discussion about these differences. In particular, I would appreciate it if you could elaborate on the relationship between the segmentation differences and the anatomical differences across the slices. 

* Minor issues

- In Table 2, column "Software," some entries state "or as open-source standalone. " I understand that this means the software can be used as a toolbox or a standalone program. Clarify that this is the meaning of that statement. Also, clarify which of the two options you used (although in principle, results should not depend on which version you used, in practice, different versions may affect the results (see, for instance, https://doi.org/10.3389/fninf.2015.00012, https://doi.org/10.1007/s00234-021-02862-0)).

- This is a matter of style, so ignore this if you disagree with this point. Change "didn't" to "did not" (line 205) and "don't" to "do not" (line 306).

- The caption of Figure 2 is split from the figure itself. Please fix this.

- Figure 4 is split between the pages. Please fix this.

- Several references to the figures are broken (line 221, line 240). Please fix them.

Author Response

Dear Reviewer,

thank you so much for your attention to our work. 

please find in the attachment our responses to your comments.

Kind regards

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

none