A Hybrid Framework for Lung Cancer Classification

Round 1

Reviewer 1 Report

I've read this article carefully, and I think it's pretty good. You want to keep improving in the classification of cancer cells, which is very good and has the spirit of scientific research, but the paper still has some shortcomings.

1.The paper lacks the network structure diagram of LCGAN and VGG-DF. This is really important. In this article, you designed LCGANT, which is composed of LCGAN and VGG-DF. The idea of LCGAN is inspired by the DCGAN. Your reference [11] shows the network structure diagram of DCGAN. As a paper focusing on network improvement, network structure diagram is indispensable. For the improved part, it is necessary to achieve graphic correspondence.

2.In the 22nd line of the text, "different carcinogens can also cause cancer", I suggest changing the expression. For example, "canceration is the result of the interaction between a person's genetic factors and three types of external factors". This is also what I got by referring to your reference [1]. Of course, you can also change the expression you are used to.

3.I don't think MoblieNet is appropriate as a comparison method. Because moblienet is a lightweight CNN network focusing on mobile terminals or embedded devices. MoblieNet has no comparative value in this paper.

4.Copying the results of other experimental methods directly and selecting Inappropriate results . In Table 4 of the paper, the evaluation indexes of other experimental methods are directly copied. you split the training and test datasets in 70:30. But you use the data that reference[31]  achieves the results  using 90% for training and 10% for testing. It will be better to reproduce other experimental methods for comparison.

I think the network structure diagram of LCGAN and VGG-DF must be added.Because there is no network structure diagram, this paper looks very incomplete.

Author Response

1. The paper lacks the network structure diagram of LCGAN and VGG-DF. This is really important. In this article, you designed LCGANT, which is composed of LCGAN and VGG-DF. The idea of LCGAN is inspired by the DCGAN. Your reference [11] shows the network structure diagram of DCGAN. As a paper focusing on network improvement, network structure diagram is indispensable. For the improved part, it is necessary to achieve graphic correspondence.

Answer:

Thanks for your suggestions, the network structure diagram of LCGAN and VGG-DF already added now.

2. In the 22nd line of the text, "different carcinogens can also cause cancer", I suggest changing the expression. For example, "canceration is the result of the interaction between a person's genetic factors and three types of external factors". This is also what I got by referring to your reference [1]. Of course, you can also change the expression you are used to.

Answer:

This is really good comment, and I followed your suggestions and changed it to:

‘Canceration results from the interaction between a person’s genetic factors and three types of external factors such as chemical carcinogens, biological carcinogens and genetic carcinogens [1]. ‘

3. I don't think MoblieNet is appropriate as a comparison method. Because moblienet is a lightweight CNN network focusing on mobile terminals or embedded devices. MoblieNet has no comparative value in this paper.

Answer:

I implemented another experiment with EfficientNet, and replaced the previous MobileNet.

4. Copying the results of other experimental methods directly and selecting Inappropriate results . In Table 4 of the paper, the evaluation indexes of other experimental methods are directly copied. you split the training and test datasets in 70:30. But you use the data that reference[31] achieves the results using 90% for training and 10% for testing. It will be better to reproduce other experimental methods for comparison.

Answer:

Thanks for your careful review! I double checked it, and removed the related experiment.

Reviewer 2 Report

See my attached report.

Comments for author File: Comments.pdf

Author Response

1. This paper needs to be polished in grammatical frame.

Answer:

Thanks for your comments. I already checked grammar  of entire essay with grammatical frame.

2. Abstract should be revised a little more.

Answer:

The abstract already revised again.

3. Figure 1 should be explained a little more.

Answer:

Thanks for your suggestion, I already add more description of Figure 1.

4. What is the physcial meanings of equation (1)?

Answer:

Thanks for your question, I added the physical meaning of equation (1).

‘Overall, the formula tries to train a discriminator which can  maximise the probability of distinguishing real images and synthetic images. It also trains a generator that can minimise the probability of distinguishing real and fake images by the discriminator. Finally, the generator can generate synthetic images that look like real images, and the discriminator cannot find the differences between them. ‘

5. Equation (3) needs to be ended with POINT.

Answer:

I added points to the Equation (3) as well as other equations.

6. This paper must be double checked in punctuation.

Answer:

Thanks for your careful review, the punctuation is double checked.

7. They need to present what is the main contribution on the studied model!

Answer:

I revised the contribution part, and represented it in more details.

8. This field of science studied previous works such as Chaos, Solitons and Fractals, 158(112050), 1-6, 2022; Comparison of compression estimations under the penalty functions of different violent crimes on campus through deep learning and linear spatial autoregressive models; Physica Scripta, 97 (054004), 1-19, 2022; Fractal and Fractional, 6(92), 1-15, 2022; Adoption of deep learning Markov model combined with copula function in portfolio risk measurement; Alexandria Engineering Journal, 61(2), 1747-1756, 2022; Modeling the pathway of breast cancer in the Middle East; Least- squares method and deep learning in the identification and analysis of name-plates of power equipment. Thus, they need to consider these works.

Answer:

Thanks for your suggestions, I carefully read these papers and cited them as the scientific background of this study.

9. The authors are requested to add more details regarding their original contributions in this manuscript.

Answer:

I added more details of our original contributions in the contribution part.

10. Papers cited in references section must be rewritten according to journal style before further process.

Answer:

Thanks for your comment. I already updated the style of the reference according to journal style.

Round 2

Reviewer 1 Report

I reread the revised paper. It looks complete. I noticed you used the LeakyReLU function. Today, the activation function is also upgrading. If you have a chance, you can try to use Sigmoid Weighted Liner Unit(SiLU) activation function. Although the paper gives the network structure diagram, the network structure diagram does not intuitively show your innovation. For example, you can use dark modules for the part of the network that has not been improved, and bright modules for the part you innovate.