Noninvasive Classification of Glioma Subtypes Using Multiparametric MRI to Improve Deep Learning
, Xinying Ren 1,2,†, Weiting Huang 3, Rui Wang 1, Laiyang Ma 1, Tiejun Gan 1, Kai Ai 4, Tao Wen 1, Yujing Li 1,2, Pengfei Wang 1, Peng Zhang 5 and Jing Zhang 1,*Round 1
Reviewer 1 Report
Dear authors,
The study appears to be scientifically sound and the discussion is well written, however the information is not well organized and individual steps are not easy to follow. I suggest that the “materials and method” section be restructured to simplify the subsections. In addition, the format needs to be improved, the manuscript has many errors.
I just have a few concerns regarding the way in which certain aspects of the study are described, as some are not provided with sufficient reference or are unclear from their descriptions. These points are described in detail below.
- The authors do not provide any reference when explaining the GradCAM in section 2.7.3. In addition, it is necessary for the authors to include some examples of the use of these method in the medical field when they said “… has been validated in the deep learning literature to assign feature importance to areas of the image”.
- The manuscript does not clarify how the data set was partitioned into training, validation, and testing. The methods section does not explain this data partition, however, Tables 2 and 3 show the test results without explaining the number of samples used in training but later in Figure 5 it does show this data partition. The authors should indicate in the section "Evaluation of the models" whether the data partition shown in Figure 5 is the same data partition used in Table 2 and 3. A summary table or graph should be added with the number of images used in each set and the number of different tumor types.
- A diagram representing the entire data flow employed in this work is necessary. This diagram should represent the image preprocessing strategies, image processing, classification based on sequence combinations and preprocessing strategies, choice of the best result, and classification with numeric data.
- Accuracy per class and overall accuracy equation must be presented in 2.7.4 section.
- In section 2.5 the use of transfer learning is not clear or not adequate for this section. An ImageNet reference is necessary, and the type of subset employed (number of images, dimension, etc).
- In section 2.5, figure 2 must be referenced before, at the end of the first sentence.
- I think that figure 5 is not correctly referenced in the manuscript.
- Results of Table 4 must be explained. It is not clear if transfer learning is applied to ResNet34 or ConvNext model. To assist the reader, a brief summary of the different methods should be added.
- It is not clear to me what figure 6 represents. What do the colors represent? Where is the incorrect prediction? ResNET34 or ConvNext model? Maybe an image in the first row pointing to the tumor may be helpful.
- The review of the state of the art in the introduction is scarce about deep learning and MRI. The authors should explain in detail what the current methods, what their limitations are, and what is being proposed in the literature.
Another minor details:
- The year of WHO CNS5 must be indicated in the Introduction section.
- The country of the hospital must be indicated.
- MR, DCNN, MRI acronym should be defined in the second paragraph.
- GradCAM in “Model Explanation” need be referred.
- In the discussion section the authors said “The GradCAM heatmaps in Figure 4”, I think that is figure 6.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors present a method for Noninvasive classification of gliomas subtypes using Multiparametric MRI.
The paper is written well, with good experimental findings. I would, however, suggest the authors include a statistical analysis of the results to establish the significance of the proposed method.
English needs to be improved.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This work is not enough contribution and innovation. However, the problem statement and motivation could be stronger or more clearly highlighted.
1. The existing literature should be classified and systematically reviewed, instead of being independently introduced one-by-one.
2. The abstract is too general and not prepared objectively. It should briefly highlight the paper's novelty as what is the main problem, how has it been resolved and where the novelty lies?
3. For better readability, the authors may expand the abbreviations at every first occurrence.
4. The author should provide only relevant information related to this paper and reserve more space for the proposed framework.
5. However, the author should compare the proposed algorithm with other recent works or provide a discussion. Otherwise, it's hard for the reader to identify the novelty and contribution of this work.
6. The descriptions given in this proposed scheme are not sufficient that this manuscript only adopted a variety of existing methods to complete the experiment where there are no strong hypothesis and methodical theoretical arguments. Therefore, the reviewer considers that this paper needs more works.
The algorithm presented has not any novelty.
7. The related works section is very short and no benefits from it. I suggest increasing the number of studies and add a new discussion there to show the advantage. Please add following:
a. Classification of Brain Tumor from Magnetic Resonance Imaging Using Vision Transformers Ensembling
b. Multimodal brain tumor classification using deep learning and robust feature selection: A machine learning application for radiologists
c. Automated detection of brain abnormality using deep-learning-scheme: A study
8. The manuscript is not well organized. The introduction section must introduce the status and motivation of this work and summarize with a paragraph about this paper.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Dear authors,
Thank you for addressing all my comments. The introduction has improved with the new references and the manuscript is better understood with the new tables and images.
