Next Article in Journal
Balancing Performance and Portability: A Study on CsI(Tl) Crystal Sizes for Real-Time Gamma-Ray Spectrum and Dose Monitoring
Previous Article in Journal
Non-Metastatic Uterine Carcinosarcoma: A Tailored Approach or One Size Fits All?
 
 
Article
Peer-Review Record

Deep Texture Analysis Enhanced MRI Radiomics for Predicting Head and Neck Cancer Treatment Outcomes with Machine Learning Classifiers

Radiation 2024, 4(2), 192-212; https://doi.org/10.3390/radiation4020015
by Aryan Safakish 1,2, Amir Moslemi 2, Daniel Moore-Palhares 2,3,4, Lakshmanan Sannachi 2,3,4, Ian Poon 3,4, Irene Karam 3,4, Andrew Bayley 3,4, Ana Pejovic-Milic 1 and Gregory J. Czarnota 3,4,5,*
Reviewer 1:
Reviewer 2: Anonymous
Radiation 2024, 4(2), 192-212; https://doi.org/10.3390/radiation4020015
Submission received: 6 May 2024 / Revised: 7 June 2024 / Accepted: 12 June 2024 / Published: 14 June 2024

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have done good job and this work will help further into radiomics-trained predictive models and DTA to boost performance and to predict cancer treatment outcomes. However the authors can work on the below comments to make the manuscript better and comprehensive.

1. Add more latest references from the field.

2. Its better if the patient outcome CR/PR is shown in graphs( like Kaplan Meier curve).

3. The resolution of the figures 3, 5, and 5 can be made better.

4. Minor English and scientific language edits needs to be done.

 

 

 

Comments on the Quality of English Language

Minor English language edits needs to be done.

 

Author Response

We thank the reviewer's for their feedback. The critique provided was helpful and in response to it the manuscript has been improved. We believe the improvements made strengthen the manuscript and look to further critique if needed.

Comment 1. Add more latest references from the field.

ReplyWe have updated the references as requested. Specifically we retain some of the older references are all linking back to original works that laid the foundation for the field (ie. Haralick et al., 1973). In order to update the references, reference [12] was changed from 2008 source to one reviewing H&N cancers from 2020, which also discussed the role of lymph node spreading for H&N cancers.

Reference [18] was from 1998 regarding HPV association with H&N updated to a modern reference (2023).

Moreover, previous reference [22] and [14] were redundant, so that issue was fixed and both are under reference [14] in updated manuscript.

The reference list was also altered to fit the MDPI format.

Comment 2. Its better if the patient outcome CR/PR is shown in graphs( like Kaplan Meier curve).

Reply: Thank you for this suggestion. We provided CR and PR data in terms of local control only, as these patient often do not die of local disease but rather concomitant metastases they have or recurrences elsewhere. In this study we correlate local control only at 3 months rather than patient survival.

Specifically, in this work we were interested in predicting the initial treatment outcomes during initial follow-up. This was selected in order to allow physicians to curate the treatment plan more effectively.

Comment 3. The resolution of the figures 3, 5, and 5 can be made better.

Reply: We have improved the resolution as requested but also in all other figures - specifically, figures 2,3,4,5, and 6 have been updated with better resolution.

Comment 4. Minor English and scientific language edits needs to be done. 

ReplyThe paper was written and has now been reviewed again by multiple first-language English speaking scientists. If there are any specific instances of English mis-usage, we will  be happy to comply with any corrections needed.

 

Reviewer 2 Report

Comments and Suggestions for Authors

hello

thank you for the paper

this is an original paper

title is sound, and OK

the paper is about MR in head and neck cancer

the title matches abstract

key words are OK

abstrac- well structured, well written

clear aim is presented

key words matches the title and abstract

introduction is sound

introduction is interesting and matches the paper title

please add a clear aim of the study at the end of this paragraph

in material and methods - please improve this chapter - clearly characterise the patients included or excluded from the study and why

add information if RTH therapy was adjuvent/neoadjuvant approach

the indications for RTH might include many scenarios - add what patients groups were studied, how and why

does the used RTH-type influence on  raiomics, RT therapy, its planning and appraocheS?

were the MR studied did on the same device and same technique? what device was it? how was it used? what reference points for RTH planning were used?

material and methods section should be divided into sub sections to more clearly describe the methods

why only 63 patients were enrolled? why the rest was not included? - improve the missing information

model section is properly presented

models are quite interesting and they represent some new features

following chapters are written OK

presented figures and tables are good and accurate

results should be more highlighted into the most important onces - please improve them

discussion is well presented

discussion should correlate with authors results

add 5 top key points at the end to highlight the most important points

conclusions could be more scientific and focusing on the most important results and their implications for future studies or RTH therapy with the usage of MR

citations should be written in MDPI style

thank you for the paper

its quite interesting, requires some revision

 

Author Response

We thank the reviewer's for their feedback. The critique provided was helpful and in response to it the manuscript has been improved. We believe the improvements made strengthen the manuscript and look to further critique if needed.

 

Comment 1.   please add a clear aim of the study at the end of this paragraph

Reply: The aim is outlined in line 142-148

 

Comment 2.   in material and methods - please improve this chapter - clearly characterise the patients included or excluded from the study and why

Reply: This has been done – we now characterize clearly why patient were included or excluded. This is further details in response for why only 63 patients enrolled.

 

Comment 3.   add information if RTH therapy was adjuvent/neoadjuvant approach

Reply: These cases were all "definitive" curative radiotherapy, not adjuvant or neo adjuvant. The "radical intent" dose was 70 Gy at least in the cases included.

 

Comment 4.   the indications for RTH might include many scenarios - add what patients groups were studied, how and why

Reply: We thank the reviewer for this comment, In reply, patients selected for study were those with bulky, pathological neck masses that were targeted for RTH such that disease could be identified and studied here using MRI but also with CT and ultrasound in companion studies. Pathological neck masses were outlined based on R.E.C.I.S.T criteria, and discussed in more detail in Section 2.2. These patients also initially recruited for another study in which we investigated the radiomic signatures from quantitative ultrasound parametric maps.

 

 

Comment:  5.  does the used RTH-type influence on  raiomics, RT therapy, its planning and appraocheS?

Reply:  Treatment was done with VMAT or IMRT methods using photon beams. We do not expect there to be any influence on the radiomics signatures as radiomic profiling takes place on pre-treatment scans. However, this is something that could be investigated in the future.

 

Comment 6.  were the MR studied did on the same device and same technique? what device was it? how was it used? what reference points for RTH planning were used?

Reply: Images were acquired with Phillips Ingenia Elition X at 1.5T field strength. Standard reference points based on CT planning were used. Isocenters are placed at the chin and shifted as needed for planning to the centre of the treatment area to ensure coverage with cone beam imaging during treatment and image verification.

 

Comment 7.  material and methods section should be divided into sub sections to more clearly describe the methods

Reply: This was done.

2.1. Treatment Approach and Follow-up Imaging

2.2. Tumour Response Definition and Segmentation

2.3. Texture Extraction and Machine Learning Algorithms

 

 

Comment 8.  why only 63 patients were enrolled? why the rest was not included? - improve the missing information

Reply:  The reason for this is because this is a subset of patients for which we also have conducted radiomics analysis with CT images, and quantitative ultrasound data. The limiting factor is the US data as the acquisition of US scans is not required for any part of the treatment pipeline, and as such, recruiting these patients was selective. From 2015-2020 we were only able to recruit ~70 patients, and not all of these had usable MR images due to various inconsistencies like dental artifacts, etc. The final number was 63.

This has now been added and it can be found in lines 168 - 171. Radiomics models for this patient cohort were trained using US and CT scans and published previously [31], [32].

 

Comment 9.   results should be more highlighted into the most important onces - please improve them

Reply: This has been done

 

Comment 10. add 5 top key points at the end to highlight the most important points

Reply: This has been added to conclusions. See line 631 – 641.

 

Comment 11. conclusions could be more scientific and focusing on the most important results and their implications for future studies or RTH therapy with the usage of MR

Reply: This has been done.

 

Comment 12 citations should be written in MDPI style

ReplyThis has been done.

 

 

Back to TopTop