CT Radiomics and Clinical Feature Model to Predict Lymph Node Metastases in Early-Stage Testicular Cancer
Round 1
Reviewer 1 Report
In this study, the Authors used radiomic features extracted from CT images, combined with clinical data, to identify patients at risk of TGCT recurrence. The study is interesting but there are several aspects that need to be clarified.
Comments
1) Radiomic features were extracted from lymph nodes. However, it is well-known that there are issues associated with the extraction and analysis of radiomic features from small volumes. Were all the used VOIs having at least 64 voxels? Did the Authors control that the investigated features were not just surrogates of VOI volume?
2) Line 170-173: the lymph nodes were randomly divided into training and test set. However, this means that lymph nodes from the same patient may be present both in training and testing set. If this was the case the two cohorts cannot be considered as independent.
3) It is not clear why Authors first split the sample into training and testing sets (line 206), but then they use 10 fold cross-validation to assess their models’ performance. Usually k fold cross-validation is used for hyperparameters tuning in the training set, or to train and test the model when sample size is relatively small. The Authors have to better explain the workflow they used and why they used it.
4) Classes were highly unbalanced. The Authors dealt with it as they could, but this should be reported as a limitation of the study anyhow.
5) Please, report (even in the supplementary material) the hyperparameters of the models used in this study
Minor comments:
1) Line 126: there is written that 91 patients were included in the analysis, then at line 139 it is reported that 76 patients were excluded. The flowchart showing the patients’ selection process is very clear, but the text must be improved. As it is right now is confusing.
2) Line 147: do not report +/- 2 weeks but include median time in days and their range.
3) Line 206: the sentence starts saying “the under-sampled dataset…” but only a few lines after it is reported that the Authors performed under-sampling. Please, be coherent. First say what you have done and then refer to it.
Author Response
Dear Reviewer,
On behalf of my coauthors, I would like to thank you for taking the time to review and comment upon our manuscript entitled “CT Radiomics and Clinical Feature Model to Predict Lymph Node Metastases in Early-Stage Testicular Cancer ” (ID: onco-2213047).
Attached we provide the point-by-point responses.
Thank you again for your thoughtful comments.
On behalf of all the co-authors
Sincerely,
Catharina Lisson
Clinician scientist, Department of Diagnostic and Interventional Radiology,
University Hospital of Ulm, Albert-Einstein-Allee 23, 89081 Ulm, Germany
(+) 49 (0) 731 500 61171
Author Response File: Author Response.docx
Reviewer 2 Report
The Authors investigated the diagnostic performance of a prediction model based on both computed-tomography based radiomic features and clinical parameters to predict for lymph node involvement in testicular germ cell tumors. The topic is of interest, methodology is robust and the manuscript is clear and well written.
Few comments:
1) Simple summary: please add low-dose adjuvant radiation therapy to the para-aortic strip as a potential treatment options.
2) Introduction: please add low-dose adjuvant radiation therapy to the para-aortic strip as a potential treatment options.
3) How did you chose the 3 retroperitoneal lymphnodes to be segment for the texture analysis?
Author Response
Dear Reviewer,
On behalf of my coauthors, I would like to thank you for taking the time to review and comment upon our manuscript entitled “CT Radiomics and Clinical Feature Model to Predict Lymph Node Metastases in Early-Stage Testicular Cancer ” (ID: onco-2213047).
Attached we provide the point-by-point responses.
Thank you again for your thoughtful comments.
On behalf of all the co-authors
Sincerely,
Catharina Lisson
Clinician scientist, Department of Diagnostic and Interventional Radiology,
University Hospital of Ulm, Albert-Einstein-Allee 23, 89081 Ulm, Germany
(+) 49 (0) 731 500 61171
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The Authors addressed all comments and the manuscript can now be considered for publication