Next Article in Journal
A Review of Trastuzumab Biosimilars in Early Breast Cancer and Real World Outcomes of Neoadjuvant MYL-1401O versus Reference Trastuzumab
Next Article in Special Issue
Towards Precision Oncology: Enhancing Cancer Screening, Diagnosis and Theragnosis Using Artificial Intelligence
Previous Article in Journal
Cancer Patients’ Experiences with Telehealth before and during the COVID-19 Pandemic in British Columbia
Previous Article in Special Issue
Applications of Artificial Intelligence in Screening, Diagnosis, Treatment, and Prognosis of Colorectal Cancer
 
 
Article
Peer-Review Record

Artificial Intelligence System for Predicting Prostate Cancer Lesions from Shear Wave Elastography Measurements

Curr. Oncol. 2022, 29(6), 4212-4223; https://doi.org/10.3390/curroncol29060336
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Curr. Oncol. 2022, 29(6), 4212-4223; https://doi.org/10.3390/curroncol29060336
Received: 21 April 2022 / Revised: 5 June 2022 / Accepted: 8 June 2022 / Published: 10 June 2022

Round 1

Reviewer 1 Report

The research aims to detect prostate cancer using machine learning models from shear-wave elastography. Three classifiers were evaluated. The experimental results demonstrated the proposed model could achieve very good accuracy and have a high potential to detect prostate cancer in clinical use. Here are some detailed comments:

 

  1. The study is lacking a comprehensive review of the recent advances of AI in cancer diagnosis. I would suggest the authors review the following papers.
    1. Wang, Haifeng, Bichen Zheng, Sang Won Yoon, and Hoo Sang Ko. "A support vector machine-based ensemble algorithm for breast cancer diagnosis." European Journal of Operational Research 267, no. 2 (2018): 687-699.
    2. Lu, Hongya, Haifeng Wang, and Sang Won Yoon. "A dynamic gradient boosting machine using genetic optimizer for practical breast cancer prognosis." Expert Systems with Applications116 (2019): 340-350.
    3. Wang, Yibin, W. Neil Duggar, Toms V. Thomas, P. Russell Roberts, Linkan Bian, and Haifeng Wang. "Extracapsular extension identification for head and neck cancer using multi-scale 3D deep neural network." In Proceedings of the 12th ACM Conference on Bioinformatics, Computational Biology, and Health Informatics, pp. 1-5. 2021.
    4. Wang, Yibin, Christian Zamiela, Toms V. Thomas, William N. Duggar, P. Russell Roberts, Linkan Bian, and Haifeng Wang. "3D Texture Feature-Based Lymph Node Automated Detection in Head and Neck Cancer Analysis." In 2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pp. 2113-2119. IEEE, 2020.
  1. It seems the parameter settings of the tested machine learning models are not mentioned. For example, the learning rate, optimizer, loss function, etc.
  2. Are any cross-validation approaches used to collect the results in table 3?
  3. The figures 4-6 can be plotted in the same figure to provide a more intuitive
  4. It seems not clear how the dataset was split into training and test.
  5. The conclusion section seems just include one sentence, which is about the impact of the study. I would suggest authors summarize the experimental results as well as the proposed models. Potential future works can also be discussed.

Author Response

Please see the attachment. Thank you!

Author Response File: Author Response.pdf

Reviewer 2 Report

Artificial intelligence system for predicting prostate cancer lesions from shear wave elastography measurements

The work has a clinical interest. After addressing some significant/major issues, described below,  I believe it could potentially be worthy of publication.

First, the article needs to be reviewed by a native English speaker to control the verb tenses that fluctuate in the draft.

  1. Question:
    Given that the sample size is relatively small and considering that the objective is testing the feasibility of the methodology of the AI proposed, the article should focus more on describing the AI technique
    proposed in great detail, highlighting its advantages and disadvantages. It is not clear what dataset has been used by the AI for training. The manuscript gives the impression that the same dataset of patients has been used when that is not right. Please give some detail on this.
  2. Question:
    In the abstract, it is indicated that the dataset is 223 patients. Is it 223x12 cores for the AI?
  3. Question:
    The introduction is relatively poor since the paper focuses on the role of AI in clinical diagnostic decision-making.
    Only one reference is cited in the entire section of AI. I think that the principles of static and dynamic elastography are well known now, and there is a large number of references that explain its beginnings, functionality, and differences. Although these techniques have been mentioned very briefly in the introduction, however, it seems that they are padding the paragraph to have a certain extension [lines 57-67] since it does not provide any relevant information to the proposed study. Please cite the references indicating the author’s name and describing a little the previous work related to prostate cancer.
  4. Question:
    In the ’methodology for elastography’ subsection, the first sentences of the second paragraph (lines 126-127) seem to talk about strain elastography. However, in the last line (132), they talk about SWE.
    Rewrite to understand if SWE (from SSI) is used or if compression is done. In other words, clarify if it is static or dynamic elastography or if both are used.
  5. Question:
    In line 124 the authors said that civco needle guide system is used for direct biopsy, but later, in line 147, the authors say that Bard Care biopsy system is used. Can the authors clarify (maybe with a schematic
    figure) which system was used? I understand that if the civco needle is attached to the SSI, it will be known where to do the biopsy since it is guided mainly in real-time by the elastography. Nevertheless,
    if the transrectal prostate biopsy is done using Bard system, how did the authors ensure that the biopsy cores corresponded precisely to the areas observed in the 2D-SWE image?
  6. Question:
    Lines 139 to 143, are not well explained. For example, what do they mean by central area and artifact? Can you please include son references from the literature? 
  7. Question:
    In subsection 3.2, what is the data set for training and testing? In what cases? Are the full 2D images part of the dataset or just the elasticity value? For individual sections?
  8. Question:
    Section 3.3 Why did the authors use these specific three machine learning classification algorithms?
  9. Question:
    Section 3.4 what does the number of layers refer to?
  10.  Question:
    What is the point of Figure 3? if the relationship type is not visible.. consider including a schematic figure.
  11. Question:
    Section 4.1 What does imbalance mean? What are the criteria to classify it as balanced?
  12. Question:
    Section 4.2 line 225, ” for our intelligent pre-screening system” What do the authors mean by this? It is clear that the AI author has written this section, but it reads unconnected to the last part of the paper.
    Lots of this section should have been explained (with higher clarity) in the methods section.
  13. Question:
    Conclusion: It should be more concise and try to conclude whether the aim was achieved or not.

Comments for author File: Comments.pdf

Author Response

Please see the attachment. Thank you!

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have improved the manuscript and have included the answers to the raised requests. However, again, I would like to ask them to include a sketch (scheme) of the process of measurements and biopsies. I think it will substantially improve the article. Furthermore, please, improve the quality and presentation of figures 3 and 4.

Author Response

Dear Reviewer,

Thank you for your suggestions, which have significantly improved our manuscript. 

According to your last recommendations, we have added a scheme of the diagnostic process, which is now Figure 3. We also have improved the quality of the last two figures, and added more information in the figure captions, to facilitate the understanding of our methodology and results.

Moreover, we have added a paragraph in the discussion section, with a few considerations regarding the advantages and drawbacks of the use of AI in practice. 

Thank you, once again, for all your invaluable help!

Back to TopTop