A Novel and Reliable Pixel Response Correction Method (DAC-Shifting) for Spectral Photon-Counting CT Imaging

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The study investigates the effects of a unique calibration/correction technique on enhancing the quality of images acquired from photon-counting systems. Overall, the job appears to be comprehensive and exceptionally well-executed.

Insignificant particulars:

To enhance comprehension, it is advisable to expand the acronym 'DAC' to its full words when it is first mentioned in the abstract, as well as in the Keywords section.

Please ensure that the phrase "et al." is written in cursive form throughout the entire manuscript.

One potential addition to the debate could involve exploring future prospects and outlining the subsequent course of action. For example, the method can be used to analyze a collection of clinical images that are assessed by radiologists in order to determine if they can detect any enhancements in image quality. Alternatively, should it be directly implemented by vendors?

Comments on the Quality of English Language

Please ensure that the phrase "et al." is written in cursive form throughout the entire manuscript.

Author Response

"The study investigates the effects of a unique calibration/correction technique on enhancing the quality of images acquired from photon-counting systems. Overall, the job appears to be comprehensive and exceptionally well-executed."

Thank you very much for your kind comments.

"Insignificant particulars:

To enhance comprehension, it is advisable to expand the acronym 'DAC' to its full words when it is first mentioned in the abstract, as well as in the Keywords section. "

Completely agreed, we have made the suggested changes in the abstract and Keywords section on page 1.

"Please ensure that the phrase "et al." is written in cursive form throughout the entire manuscript. "

Thank you for the suggestion, after reviewing the last five papers published in MDPI Tomography, we found that none of them used italics for "et al." and the practice is supposedly becoming less common. We will therefore leave it as is to maintain consistency with the journal.

"One potential addition to the debate could involve exploring future prospects and outlining the subsequent course of action. For example, the method can be used to analyze a collection of clinical images that are assessed by radiologists in order to determine if they can detect any enhancements in image quality. Alternatively, should it be directly implemented by vendors?"

Thank you for the excellent points. In the next steps we would like to apply our imaging system to more relevant clinical questions. In our research, we focus on high resolution tumor imaging of breast cancer patients. For that specific application we think it is mandatory to mitigate the challenges of pixel-response variation, and given the instability of our system, DAC-shifting seems to be the best option. From the vendor perspective, it mainly depends on what the vendors are selling. Currently you will most likely buy a detector from a vendor, and as it is a post-processing method, it is less likely that vendors will provide such a method within their software. Alternatively, if you buy a complete system, like the MARS bioimaging system, it could be easier to implement up front. We think these discussion points are interesting, but go beyond the scope of the current paper. 

"Please ensure that the phrase "et al." is written in cursive form throughout the entire manuscript."

See previous related comment.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript titled "A novel and reliable pixel response correction method (DAC-shifting) for spectral photon counting CT imaging" presents a new post-processing method called DAC-shifting to address the challenges of individual pixel response behavior in spectral photon counting cone-beam CT imaging. The authors have developed this method to correct projection data by explicitly measuring pixel response using flat field images. They compared DAC-shifting against the signal-to-thickness calibration (STC) method and found it more effective, particularly in scenarios with detector instability and over time.

The paper is well-structured, with clear sections detailing the introduction, methods, experimental setup, and results. The authors provide a comprehensive overview of the problem with current pixel response correction methods and offer a detailed explanation of the DAC-shifting method. They validated their method using phantom imaging and K-edge imaging, demonstrating its effectiveness in improving image quality and consistency in spectral CT imaging. The manuscript also includes detailed tables and figures that support their findings. The study's significance is underscored by the potential impact on clinical imaging, particularly in enhancing the accuracy of material decomposition and contrast agent differentiation.

However, further details about the methodology and experimental setup would enhance the clarity and depth of the study. From the contributions the authors described in the manuscript, I observe a deficiency in the requisite level of novelty or originality that would warrant publication in its current form and suggests exploring alternative venues. The authors should refine the contributions of this work.

In addition, there are several typo issues and writing errors. Some necessary details about the proposed technique are missing.

Below are some detailed comments:

Major concerns:

1. The contributions are not listed in current version. It would be beneficial for the authors to refine and list their contributions, aiding the audience in grasping the key elements of this work.

2. The Discussion Section, being a crucial element of a scientific paper, appears to be lacking in depth. It would be beneficial for the authors to significantly enhance its content.

3.     I suggest restructuring the manuscript, in particular, the authors should list the limitations of this work, move the corresponding content in the "Discussion" section to "Introduction." This would improve the overall flow and clarity of the paper.

4.     Some of the Figure captions or Table captions can be simplified, e.g., Figures 2, 6 and Table 4.

5. There is a lack of real ‘Discussion,’ The current Discussion section resembles an extension of the Introduction.

Minor comments:

1.     Spell out " PMMA ", “STC”, “DAC-shifting” when they’re first used in the Abstract section.

2.     There are some grammar/writing issues.

3.     Please double check all the cited references in the Section of References, making all the references in the same format. Consider adding more recent studies if available.

4.     To facilitate reproducible research in scientific works, it is highly desirable to provide access to the code and exemplar data to support experiments described in the manuscript.

Comments on the Quality of English Language

The quality of English Language need to be improved

Author Response

"The manuscript titled "A novel and reliable pixel response correction method (DAC-shifting) for spectral photon counting CT imaging" presents a new post-processing method called DAC-shifting to address the challenges of individual pixel response behavior in spectral photon counting cone-beam CT imaging. The authors have developed this method to correct projection data by explicitly measuring pixel response using flat field images. They compared DAC-shifting against the signal-to-thickness calibration (STC) method and found it more effective, particularly in scenarios with detector instability and over time.

The paper is well-structured, with clear sections detailing the introduction, methods, experimental setup, and results. The authors provide a comprehensive overview of the problem with current pixel response correction methods and offer a detailed explanation of the DAC-shifting method. They validated their method using phantom imaging and K-edge imaging, demonstrating its effectiveness in improving image quality and consistency in spectral CT imaging. The manuscript also includes detailed tables and figures that support their findings. The study's significance is underscored by the potential impact on clinical imaging, particularly in enhancing the accuracy of material decomposition and contrast agent differentiation.

However, further details about the methodology and experimental setup would enhance the clarity and depth of the study. From the contributions the authors described in the manuscript, I observe a deficiency in the requisite level of novelty or originality that would warrant publication in its current form and suggests exploring alternative venues. The authors should refine the contributions of this work.

In addition, there are several typo issues and writing errors. Some necessary details about the proposed technique are missing."

Thank you very much for your detailed feedback, we take your concerns seriously. As these comments are detailed below, we address them there.

"Below are some detailed comments:

Major concerns:

1. The contributions are not listed in current version. It would be beneficial for the authors to refine and list their contributions, aiding the audience in grasping the key elements of this work."

In lines 80-88 in the introduction we introduced the work, describing that we developed a novel post-processing method that does not need calibration measurements. With this method we achieved robustness agains system instabilities, and we achieve freedom in choosing x-ray source settings, and in selecting the number and level of the spectral thresholds. In some journals it is common to specifically highlight this in bullet points in the introduction section, but this is not common for the Tomography journal. In our opinion, the novelty of the paper can be clearly deducted from the text in the abstract, and in the introduction. This is also supported by the fact that the other reviewers did underline the novelty of the approach.

"2. The Discussion Section, being a crucial element of a scientific paper, appears to be lacking in depth. It would be beneficial for the authors to significantly enhance its content."

It is unclear to us what is ment by "lacking in depth". We have written 109 lines of discussion, including all the relevant literature that we could find to put the results of the work into perspective of the existing knowledge. Unfortunately, there are not many papers available that have conducted research where the stability over time is assessed. This is one of the strong points of our method, and it would have been nice if we could have compared results with others. If you could provide specifics of what you are missing in the discussion, we are of course willing to further adapt the discussion. 

Based on the comments of reviewer 3, we did expand our section on the future use of neural networks to perform the DAC-shifting method (Lines 435 - 439), and we also added a section on the wider application of the method towards other sensor materials and detector types (Lines 440-446)

3. I suggest restructuring the manuscript, in particular, the authors should list the limitations of this work, move the corresponding content in the "Discussion" section to "Introduction." This would improve the overall flow and clarity of the paper.

We agree that limitations should also be mentioned in the paper, as we have done in the discussion section. However, we disagree that this section should be moved to the introduction part. In our opinion, this would actually make the paper more confusing, as the limitations can not be fully understood before the reader has been through the materials and method section, and the results. We also think that limitations at the end of the discussion is common practice in the Tomograpy journal

"4. Some of the Figure captions or Table captions can be simplified, e.g., Figures 2, 6 and Table 4."

This is indeed quite a technical and detailed paper, we have revised the aforementioned captions. Please find the updated captions in Figures 2, 6 and Table 4.

5. There is a lack of real ‘Discussion,’ The current Discussion section resembles an extension of the Introduction.

Similar to point 2 above, it is unclear to us what is ment by "real Discussion". In our opinion, the introduction and discussion need to be cohesive, as the introduction should clearly state what is lacking in existing methods, and the discussion should contain an explenation about how the new work is filling that gap. 

Minor comments:

"1. Spell out " PMMA ", “STC”, “DAC-shifting” when they’re first used in the Abstract section."

Corrected

"2. There are some grammar/writing issues."

We have performed a spelling check, and had a native English speaker check the grammer.

"3. Please double check all the cited references in the Section of References, making all the references in the same format. Consider adding more recent studies if available."

References 16 and 17 appear to be the only references of a different format to the rest. This is because the first is a website and the second is a PhD thesis. Therefore, different formats must be used compared to the rest, which are all papers.

We have also updated our literature search, and did not identify new relevant papers for the existing part of the text. For the additions in the discussion we also added 3 new references [25, 26, and 27]

"4. To facilitate reproducible research in scientific works, it is highly desirable to provide access to the code and exemplar data to support experiments described in the manuscript."

In our "Data Availability Statement" (page 23) it states: "Imaging data that was used in this study can be shared upon request." There is in principle no barier for sharing of the code, except for the time needed to make it general enough for external use. Our code has been optimized to run on the available hardware, and is not easily portable on other devices. We prefer to focus our time on further developing the method into a deep-learning solution, and will be happy to share the network at a later stage.  

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript describes DAC-shifting to improve pixel response in spectral photon counting CT imaging. The authors showed that the method is advantageous compared with traditional methods. But, the paper falls short in several areas.

1. The author used the Medipix3 chip in their research. However, there are several competing chips available on the market, such as the PILATUS series, the Rayence C-Series, and the Timepix3 series. The author needs to consider whether the research findings are generalizable or if the results are specific to the Medipix3 chip.

2.The paper contains an excessive number of figures, some of which provide limited support to the main theme. ⁤⁤The author can move some of them to the appendix. ⁤ ⁤

3. The discussion section lacks clarity on how neural networks enhance computational speed. ⁤⁤Could the author elaborate on the specific mechanisms involved? ⁤

⁤4. The abstract and conclusion should provide quantitative data to substantiate the claims of 'reducing ring artifacts' and 'improving image quality'. ⁤

Author Response

"The manuscript describes DAC-shifting to improve pixel response in spectral photon counting CT imaging. The authors showed that the method is advantageous compared with traditional methods. But, the paper falls short in several areas."

"1. The author used the Medipix3 chip in their research. However, there are several competing chips available on the market, such as the PILATUS series, the Rayence C-Series, and the Timepix3 series. The author needs to consider whether the research findings are generalizable or if the results are specific to the Medipix3 chip."

This is a good point, we expect that any photon counting chip that can change energy threshold will be able to use this method. Specifically, the PILATUS detectors should be compatible with this. The Rayence C-Series are flat panel detectors and so are not the correct detector type to use this method. For Timepix3, there are two reasonable approaches. The first would be to use it in frame based readout mode and then this method makes sense. The other option is to calibrate each pixel individually using ToT (time over threshold) and ToA (time of arrival) information, this is the typical approach we have seen in high energy physics contexts. A similar approach could be used for x-rays.

We have added "In this work, we have only used Medipix3 detectors with a silicon sensor, but the method can in principle be used in any type of spectral photon counting detector and sensor material. Sensor materials with k-edges in the region of scanning may introduce complications, this was not investigated in this work. The minimum requirement is that multiple spectral thresholds need to be acquired, including flat-field images. Example detectors could be the Dectris PILATUS series [26], and also the Timepix family used in frame based readout or time-over-threshold mode  [27]." In lines 440 to 446.

"2.The paper contains an excessive number of figures, some of which provide limited support to the main theme. ⁤⁤The author can move some of them to the appendix. ⁤ ⁤"

We understand your point, there are many figures indeed. However we feel that the current set of figures is required to adequately illustrate our method in the detail we desire. As the journal is online only, we feel that it is just as convenient to have all the information centralized, instead of moving information to supplementary material.

"3. The discussion section lacks clarity on how neural networks enhance computational speed. ⁤⁤Could the author elaborate on the specific mechanisms involved? ⁤"

Indeed the content on neural networks is brief currently. We have elaborated on the specific mechanisms involved.

We have added "The neural network then needs to perform the spectral fitting, and determine the corrections for the projection data. This can most likely be done using a physics-informed neural network [ 25 ], where we can use the raw and DAC-shifted images for supervised learning. The speed-up then comes from highly parallelised GPU processing of the data." in lines 435 - 439.

"⁤4. The abstract and conclusion should provide quantitative data to substantiate the claims of 'reducing ring artifacts' and 'improving image quality'. ⁤"

We have added quantitative data in the both the abstract and the conclusion section.