Dual-Tracer PET Image Separation by Deep Learning: A Simulation Study
Round 1
Reviewer 1 Report
The authors propose a deep learning method to separate the images of two PET traces that are administered within the same imaging session. The manuscript is well written and the methodology and results are appropriately accomplished. Discussion could be extended as I see many questions and uncertainties to be able to transfer this technique to real patients. I believe the number of figures could be much reduced without lost of information.
Would DL-method be valid for other tracers? for example with 2 tracers based on F18? would it be valid for more that two tracers? could you apply DL-method on sinogram data?
There are only minor variations on the kinetic constants introduced in the simulated images. To what extend this methodology would be valid for an extended patient variability where different kinetic constants might appear. Let's say for example for healthy patients or for different degree of diseases.
To what extend your results might be affected by the chosen framing scheme? The shortest frames are 15 seconds. Isn't that a bit long for kinetic studies, especially for the first minutes where fast variation occurs?
Could you recover the kinetic constants following a similar strategy?
The authors perform simulations based on a numerical phantom including randomized structures and tumor locations. However, brain contour and most brain structures might be the same for all cases. Furthermore, only 3 different homogenous regions are considering including white and gray matter and the tumor. Finally, only separated 2D images are considered. In order to apply the proposed methodology to patients, 3D images which much more heterogeneity would be faced. To what extend the authors believe that DL-method could be used in patient data?
The link provided for the DIRECT package is not working. Some information about this package would be useful.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The conclusion must need to summarize research study
Comparative analytics can improve manuscript
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
· Authors should explain the reason why they choose these methods. What are the limitations of this work? How can the rigor of this work be demonstrated?
· Authors should add more details about the implementation of the code to perform the analysis and the library involved in this task.
· Authors should add the parameters of the methods.
· It would be better to add some necessary arguments for Equations to make them easier to understand.
· What is the motivation of the proposed work? Research gaps, objectives of the proposed work should be clearly justified. The authors should consider more recent research done in the field of their study. Such as Simultaneous reconstruction and segmentation of dynamic PET via low-rank and sparse matrix decomposition. Simultaneous reconstruction and segmentation for dynamic SPECT imaging.
· Please use a simple diagram or figure to illustrate the whole idea of this paper, and the modification it has been made from previous work or traditional framework.
· Experimental results are not clear. What are the parameters used in the proposed system and how their values are set? Also, how the parameter values can affect the proposed system? Sections like Experimentation have to be extended and improved thus providing a more convincing contribution to the paper
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
The manuscript was prepared in high quality. The authors presented a nice overview of the literature background and scientific task. They also properly explained their suggestion, the deeplearning-based method as alternative way to evaluate the dual tracer PET imaging simultaneously compared to the well known compartment modelling. Excellent work, I suggest to publish without any correction.
Author Response
We would like to thank the anonymous reviewer for his/her very detailed reviews of our work and for the very helpful remarks.
Round 2
Reviewer 1 Report
I acknowledge the effort made by the author to reply to my comments and to improve the manuscript.
Reviewer 3 Report
none
