Deep Learning-Based Delayed PET Image Synthesis from Corresponding Early Scanned PET for Dosimetry Uptake Estimation
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors developed a deep learning model based on conditional generative adversarial network (cGAN) Image-to-image translation (I2I) methodology to generate synthetic delayed PET image from early scans of PET image and estimate dosimetry uptakes. 18 healthy participants imaging at different timepoints are analyzed. Later timepoint images are generated by the deep learning tool. SUV values are calculated and are comparable in muscle, heart, liver and spleen to the ground-truth values of actually later timepoint images. Although, SUV values have discrepancies in brain and kidney. This should be improved if more participants’ data are involved and analyzed. As the authors mentioned, cancer patients’ image data should be considered to improve this learning model in future to address the limitation of this study. To the best of my knowledge, only few attempts have been reported for the deep learning synthesis of PET images. This study developed the deep learning model that could predict the later timepoint image based on the early scans that can potentially save time in terms of dosimetry calculation or time activity curve generation. I recommend this manuscript to be published in Diagnostics with minor revision (see attached PDF file).
Comments for author File: 
Comments.pdf
Decent English whild could be further polished by native speakers.
Author Response
|
Response to Reviewer 1 Comments |
||
|
1. Summary |
|
|
|
Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files. |
||
|
2. Questions for General Evaluation |
Reviewer’s Evaluation |
Response and Revisions |
|
Does the introduction provide sufficient background and include all relevant references? |
Yes |
|
|
Are all the cited references relevant to the research? |
Yes |
|
|
Is the research design appropriate? |
Yes |
|
|
Are the methods adequately described? |
Yes |
|
|
Are the results clearly presented? |
Yes |
|
|
Are the conclusions supported by the results? |
Yes |
|
|
3. Point-by-point response to Comments and Suggestions for Authors |
||
|
Comments 1: I recommend this manuscript to be published in Diagnostics with minor revision (see attached PDF file). |
||
|
Response 1: Thank you for pointing this out. We agree with this comment. Therefore, we have corrected all expressions you pointed out as followings. Changing ‘demonstrates’ to ‘demonstrate’ in line 11 in abstract, page 1, ‘get’ to ‘getting’ in line 10, page 1, ‘the’ to ‘this’ and ‘PET’ to ‘PET image’ in line 13, page 2, ‘argue’ to ‘argued’ in line 2, page 4, and ‘organs’ to ‘spleen’ in line 38, page 9. We also have removed unnecessary period in line 18, page 9. |
||
Reviewer 2 Report
Comments and Suggestions for AuthorsThank you for submitting this paper for my review. I appreciate you allowing me to provide feedback, and hope you find my comments constructive.
I would recommend expanding your results to include metrics characterizing the distribution of SUV errors between the real and synthetic PET scans, rather than only reporting the mean error per organ. Specifically, it would be useful to include the standard deviation, minimum/maximum, median, and percentage within acceptable range. This would provide fuller insight into the variation and spread of errors across patients. With only 2 test patients, the mean can be misleading if some errors are quite large. More robust error statistics would strengthen your reported performance. Additionally, further discussion of the clinical significance and potential real-world impact of this technique would enrich the manuscript. Elaborating on practical applications along with limitations and future directions would highlight the significance of your work.
Overall this is solid research with potential for impact. Thank you again for the opportunity to review. I encourage you to continue advancing this line of work, and look forward to seeing your future innovations in the field.
Author Response
|
Response to Reviewer 2 Comments |
||
|
1. Summary |
|
|
|
Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files. |
||
|
2. Questions for General Evaluation |
Reviewer’s Evaluation |
Response and Revisions |
|
Does the introduction provide sufficient background and include all relevant references? |
Yes |
|
|
Are all the cited references relevant to the research? |
Yes |
|
|
Is the research design appropriate? |
|
|
|
Are the methods adequately described? |
Yes |
|
|
Are the results clearly presented? |
Can be improved |
|
|
Are the conclusions supported by the results? |
Can be improved |
|
|
3. Point-by-point response to Comments and Suggestions for Authors |
||
|
Comments 1: I would recommend expanding your results to include metrics characterizing the distribution of SUV errors between the real and synthetic PET scans, rather than only reporting the mean error per organ. Specifically, it would be useful to include the standard deviation, minimum/maximum, median, and percentage within acceptable range. This would provide fuller insight into the variation and spread of errors across patients. With only 2 test patients, the mean can be misleading if some errors are quite large. More robust error statistics would strengthen your reported performance.
|
||
|
Response 1: Thank you for pointing this out. We agree with this comment. Therefore, we have modified the description on the SUVmean estimation in section 3.2, page 7, and the Figure 5 in page 8 to specify the results. Previous results was only for a patient in test dataset and the description was too vague. Hence, we added related explanation. Moreover, we considered that calculating statistics for only two patient can be misleading as you pointed out. Indeed, we found that the variation of SUVmean in brain was quite large between patients. Therefore, we plotted patient-wise data in Figure 5.
|
||
|
Comments 2: Additionally, further discussion of the clinical significance and potential real-world impact of this technique would enrich the manuscript. Elaborating on practical applications along with limitations and future directions would highlight the significance of your work.
|
||
|
Response 2: Thank you for pointing this out. We agree with this comment. Therefore, we have added the discussion on the clinical significance and its potential in page 9 and modified the arrangement of the paragraphs in the discussion. |
||
