Image Quality Improvement in Deep Learning Image Reconstruction of Head Computed Tomography Examination
Round 1
Reviewer 1 Report
GENERAL COMMENTS
Throughout the text, commas must have somehow been replaced by full stops.
The English language at some points throughout the text needs sme editing.
2. Materials and Methods
Lines 57-64: please add more details about the acquisition protocol (for example range of mA).
Lines 87-88: please replace “ … the mean attenuation value (mean) …” with “ … the mean CT number …”
Figure 1: The caption of Figure 1 is cropped. Why are the two sets of images presented separately?
Lines 91-92: please rephrase, it would suffice to state the equations for SNR and CNR
Lines 103-104: what is the reason for selecting p<0.01 for the level of statistical significance instead of the usual p<0.05?
Statistical analysis: did the authors perform a statistical test to check the normality of their data? Such a test is necessary to justify the applicability of paired samples t-test instead of the non-parametric equivalent.
3. Results
Most of the paragraph either repeats information already mentioned in the Materials and Methods section (first 2 sentences) or presents comments that would better fit in the Discussion section. To make things worse, these comments are presented before the actual results. In my opinion the whole section must be fully rewritten, in order to just present the results from the objective (Table 1) and the subjective evaluations. For the latter, I believe that another table or graph would help the readers apprehend better the results.
Lines 119-120: The table does not present “All SNR and CNR measurements … ” instead I suppose it presents mean values.
Table 1: please add a title to the table. In my opinion, it would be useful to add a column with the mean and the standard deviation of the differences between DLIR and ASIR, for each parameter. Since all p values were calculated <0.001, the last column can be replaced by a footnote, indicating that.
4. Discussion
Line 133: what do the authors mean by “HU volumes”?
Lines 137-139: The values for the increase in SNR and CNR do not match those presented in Table 1. Especially increases in CNR greater than 150% are mentioned here, whereas Table 1 shows increases of the order of 50% !
Lines 146-158: This whole paragraph is not commenting on the results, most of its content fits better to the Introduction section.
Some improvement of the English language seems necessary.
Author Response
Please see the attachment
Author Response File: Author Response.docx
Reviewer 2 Report
It is too difficult for me to review this manuscript in its current form.
The manuscript exhibits significant grammatical errors, most notably the confusion between commas and periods. This pervasive issue makes the paper exceedingly challenging to read and comprehend. Please introduce abbreviations then they are first used in the manuscript.
Author Response
The authors are deeply sorry for the inconvenience of commas switching to full stops the whole text. The authors do believe they fixed the problem that must have appeared during file conversion and hope that paper is suitable for review in its current form.
Round 2
Reviewer 1 Report
No further comments
Author Response
Thank you for the provided support.
Reviewer 2 Report
Dear Authors,
This study investigates the application of deep learning image reconstruction (DLIR) in non-contrast head CT examinations, which is a promising area of research. However, there are several points that I believe need to be addressed to strengthen the manuscript.
2. Purpose of the study: The manuscript does not clearly state the purpose of the study. It would be beneficial to clarify why reconstructing head CT using DLIR vs ASiR is warranted and how this study contributes to that goal specifically in your area of clinical interest. What are the objectives of your study exactly? As DLIR is a published and commercially available algorithm what are you contributing with this study?
3. Novelty: The novelty of the study is not clearly articulated. It appears that the study is applying commercially available software to enhance the visibility of CT scans. If there were modifications or improvements to the DLIR reconstruction algorithms of TrueFidelity (GE Healthcare deep neural network engine), these should be clearly stated.
4. If this is a head-to-head comparison of two methods then this study should be structured in that way where ASiR is the gold standard and DSIR the new method. Authors should then provide a Bland Altman analysis to provide insight on observed differences and possible biases. A paired t-test should not be applied uncritically to method comparison data. Only when the graphical display suggests that a systematic constant difference, but not a systematic proportional difference, is involved should this test be applied.
5. Evaluation by radiologists: The evaluation was conducted by three radiologists with different levels of expertise. It would be beneficial to provide a justification for the number of evaluators and how the final values were obtained (majority vote?). Additionally, it would be helpful to discuss how the variation in expertise was considered in the evaluation. Please provide agreement statistics.
6. Level of reconstruction and evaluation: It is not clear whether the reconstructions and evaluations were conducted for every CT slice or the entire CT scan. Please clarify this point. If it was at the slice-level, how were the slices chosen? If it was at the scan-level, was the evaluation of imaging quality averaged over all slices? Please discuss the reasoning behind this choice.
There are some abbreviations, such as DLIR, that are used in the abstract without prior definition (see abstract). It would be helpful for readers if these abbreviations were defined at their first use (using correct naming) and only once per section. There are still some minor typos, specifically confusion of comma with period. (e.g., line 90, line 178). Please check English grammar in the entire manuscript again.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 3
Reviewer 2 Report
Thank you for your reviewed manuscript. It is much improved and I only have a few minor comments:
1- Please include in your methods a description of your BA analysis and a brief statement in your results (what you provided in your response letter). You should report the LOA and confidence limits as well as the bias (mean difference). The graphs can be included in supplementary material or in the paper (my preference). This is important as you are comparing methods and the BA analysis is the best way for readers to appreciate the difference between the methods.
2- I appreciate that you essentially used expert assessment in order to label your images. However, as you involved 3 readers it would be important to report how much they were in agreement before the senior reader made the final decision. This can be reported as Fleiss' Kappa.
There needs to be a final review of the grammar before print.
Author Response
Please see the attachment.
Thank you sincerely for all of the provided advice and support.
Author Response File: Author Response.docx