Lung Cancer Prediction Using Robust Machine Learning and Image Enhancement Methods on Extracted Gray-Level Co-Occurrence Matrix Features
, Hadeel Alsolai 3, Siwar Ben Haj Hassine 4, Mohamed K. Nour 5, Mesfer Al Duhayyim 6, Anwer Mustafa Hilal 7,*, Ahmed S. Salama 8, Abdelwahed Motwakel 7, Ishfaq Yaseen 7 and Mohammed Rizwanullah 7
Round 1
Reviewer 1 Report
Authors proposed methodology to detect lung cancer prediction. After reading submitted manuscript, my comments are as follows :
- In title, authors mention the extraction of features from GLCM, but in abstract as well as in submitted manuscript, there is no description of GLCM technique. Kindly include necessary description in revised manuscript.
- Introduction portion should be more literature oriented and need more discussion.For Ex : GLCM related literature (multidisciplinary) etc.
- Section 2.1 include the information about dataset used by authors to conduct study. Dataset description should be enhanced with more details.
- Description about ML algorithms like SVM, Naive Bayes, Decision trees should be short as these are well established algorithms.
- Quality of bar charts in Fig.4 and Fig.5 should be improved, similar suggestions for other figures also.
- It is suggested to include literatures related to GLCM in revised manuscript : a. https://www.mdpi.com/2076-3417/12/8/3715. b. https://www.mdpi.com/2073-8994/14/2/236. c. http://scindeks.ceon.rs/article.aspx?artid=1451-20922002468P
- Suggest to compare results with in Table 1 with recently published literatures (2020,2021 etc.)
Author Response
Reviewer # 1:
Comment 1
- In title, authors mention the extraction of features from GLCM, but in abstract as well as in submitted manuscript, there is no description of GLCM technique. Kindly include necessary description in revised manuscript.
Response 1
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 29-30, 171-188.
Comment 2
- Introduction portion should be more literature oriented and need more discussion.For Ex : GLCM related literature (multidisciplinary) etc.
Response 2
The issue has been addressed
Comment 3
- Section 2.1 include the information about dataset used by authors to conduct study. Dataset description should be enhanced with more details.
Response 3
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 85-90.
Comment 4
- Description about ML algorithms like SVM, Naive Bayes, Decision trees should be short as these are well established algorithms.
Response 4
The issue has been addressed
Comment 5
- Quality of bar charts in Fig.4 and Fig.5 should be improved, similar suggestions for other figures also.
Response 5
The issue has been addressed
Comment 6
- It is suggested to include literatures related to GLCM in revised manuscript : a. https://www.mdpi.com/2076-3417/12/8/3715. b. https://www.mdpi.com/2073-8994/14/2/236. c. http://scindeks.ceon.rs/article.aspx?artid=1451-20922002468P
Response 6
The issue has been addressed see highlighted lines 171-188 in red color
Comment 7
- Suggest to compare results with in Table 1 with recently published literatures (2020,2021 etc.)
Response 7
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 338-339
Author Response File: 
Author Response.docx
Reviewer 2 Report
Page 2, Paragraph 1, Line 3, correct correct the number (delete 2021).
Page 2, Paragraph 1, Line 5, where you wrote "among those", did you mean "among others"? May need correction
Page 4, equation (1): what do represent d, c?
Author Response
Reviewer # 2:
Comment 1
Page 2, Paragraph 1, Line 3, correct correct the number (delete 2021).
Response 1
The issue has been addressed
Comment 2
Page 2, Paragraph 1, Line 5, where you wrote "among those", did you mean "among others"? May need correction
Response 2
The issue has been addressed
Comment 3
Page 4, equation (1): what do represent d, c?
Response 3
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 115, 141.
Author Response File: Author Response.docx
Reviewer 3 Report
This work utilized different image enhancement methods to improve the lung cancer prediction using CT scan. Different machine learning based classifications are used to evaluate the classification accuracy.
Major comments:
1. The authors declared in ‘Chapter 1. Introduction’ that ‘We proposed image enhancement methods such as image adjustment, gamma correction, thresholding, contrast stretching to improve the quality of input images, noise removal etc.’ However, none of those methods are novel, all of them have been widely used in medical image processing systems. And the classification algorithms applied in the manuscript are also not new. The contribution of the work is limited.
2. According to the evaluation results, it is very likely that the classifier was overfitting, more data is needed to verify the performance of those methods.
3. The evaluation results should include not only the average values but also their standard deviation.
Minor comments:
1. There are extra white line and cursor in Figure 1 and the resolution of subfigure (a) – (e) is too low to read clearly.
2. What does ‘d’ and ‘c’ mean in Equ (1)?
3. Please align the equations, e.g., Equ (9), (10), (11).
4. Please make the font in each figure consistent
5. Please double check all the typos, e.g.,
We the extracted the texture features -> then
Author Response
Reviewer # 3:
Comment 1
- The authors declared in ‘Chapter 1. Introduction’ that ‘We proposed image enhancement methods such as image adjustment, gamma correction, thresholding, contrast stretching to improve the quality of input images, noise removal etc.’ However, none of those methods are novel, all of them have been widely used in medical image processing systems. And the classification algorithms applied in the manuscript are also not new. The contribution of the work is limited.
Response 1
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 68-83
Comment 2
- According to the evaluation results, it is very likely that the classifier was overfitting, more data is needed to verify the performance of those methods.
Response 2
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 88-93, 368-376.
Comment 3
- The evaluation results should include not only the average values but also their standard deviation.
Response 3
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 368-376.
Comment 4
- There are extra white line and cursor in Figure 1 and the resolution of subfigure (a) – (e) is too low to read clearly.
Response 4
The issue has been addressed
Comment 5
- What does ‘d’ and ‘c’ mean in Equ (1)?
Response 5
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 118
Comment 6
- Please align the equations, e.g., Equ (9), (10), (11).
Response 6
The issue has been addressed , equations removed as suggested by reviewer 1.
Comment 7
- Please make the font in each figure consistent
Response 7
The issue has been addressed
Comment 8
. Please double check all the typos, e.g.,
We the extracted the texture features -> the
Response 8
The issue has been addressed, the whole manuscript is edited for English
Author Response File: Author Response.docx
Reviewer 4 Report
Reviewer’s Report on the manuscript entitled:
Image Enhancement methods on extracted Gray-level Co-occurrence Matrix (GLCM) based features to improve the Lung Cancer prediction by applying robust machine learning Techniques
The authors attempted to improve lung cancer image quality by utilizing and employing various image enhancement methods such as image adjustment, gamma correction, contrast stretching, thresholding, and histogram equalization methods. Though the topic is a crucial topic in the field of medicine, the manuscript is poorly written and requires extensive revisions. Below please see my comments.
Please note that once you prepare the revision, you should insert the line numbers!
The title of the manuscript is too long and should not contain abbreviations. A suitable title could be:
Lung cancer prediction using robust machine learning and image enhancement methods
Line 3 in the abstract. Please remove “the” before researchers.
Line 7 in the abstract. “We the…”?
In the abstract, you say 100% accuracy. Before this sentence, you should mention what type of datasets you used that you achieved this accuracy because the current way of writing gives the impression to the readers that your method is perfect for all types of datasets which is not true. So write something like the following, “Our analysis on 945 images provided by the Lung Cancer Alliance
Figure 1. Please enlarge the font size in panels (a)-(d). Instead of SVM Poly., write the full name in that box: SVM Polynomial
This is very important. The font size of all the texts and numbers in all the figures (labels, axis values, legends, texts, etc.) should all have the same size and be approximately the same size as the font size of the figure caption.
Figures 3 and 6. There is a lot of blank space in the panels. Figure 6 can be removed and the instead the results can be briefly summarized in a table. Figure 3 can also be improved. You may start the y-axis from 0.5 instead of 0 so that the blank spaces can be reduced.
Figure 1 and its description should be moved to the method section. At the end of Introduction please mention how the rest of the manuscript is organized by referring to the section numbers.
In section 2.1 Define DICOM, SCLC, etc. Please note that all the abbreviations should be defined the first time they are used in the manuscript. Please also add an acronym table at the end of the manuscript.
The link you provided www.giveascan.org looks broken. Please mention the last access date. Also, you need to put all the links in the References and treat them as regular references.
Section 2.4 after “…content-based applications” please add the following articles:
https://doi.org/10.1109/SPIN.2017.8049930
https://doi.org/10.3390/s22082948
https://doi.org/10.3390/s22062346
Before Equation (12), inside the {…}, please replace “a_i,” with “a_N”. Please carefully check all the equations.
Page 9. Please carefully check whether the values in percentage match the one in Figure 2 when you described them.
Please write the limitations of the study and provide future direction in the conclusion part as well.
Please carefully check the references to ensure they are correct and have a consistent format according to the MDPI guidelines.
Thank you!
Regards,
Author Response
Reviewer # 4:
Comment 1
The title of the manuscript is too long and should not contain abbreviations. A suitable title could be:
Lung cancer prediction using robust machine learning and image enhancement methods
Response 1
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 1-3
Comment 2
Line 3 in the abstract. Please remove “the” before researchers.
Response 2
The issue has been addressed
Comment 3
Line 7 in the abstract. “We the…”?
Response 3
The issue has been addressed
Comment 4
In the abstract, you say 100% accuracy. Before this sentence, you should mention what type of datasets you used that you achieved this accuracy because the current way of writing gives the impression to the readers that your method is perfect for all types of datasets which is not true. So write something like the following, “Our analysis on 945 images provided by the Lung Cancer Alliance
Response 4
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 35-36.
Comment 5
Figure 1. Please enlarge the font size in panels (a)-(d). Instead of SVM Poly., write the full name in that box: SVM Polynomial
Response 5
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 97-98.
Comment 6
This is very important. The font size of all the texts and numbers in all the figures (labels, axis values, legends, texts, etc.) should all have the same size and be approximately the same size as the font size of the figure caption.
Response 6
The issue has been addressed
Comment 7
Figures 3 and 6. There is a lot of blank space in the panels. Figure 6 can be removed and the instead the results can be briefly summarized in a table. Figure 3 can also be improved. You may start the y-axis from 0.5 instead of 0 so that the blank spaces can be reduced
Response 7
The issue has been addressed. The suggestion regarding Figure 6 are highly appreciated, however, usually fold 1 to 10 visual ROC is usually preferred, we changed the y-axis 0.5 to 1 for Figure 6 too.
Comment 8
Figure 1 and its description should be moved to the method section. At the end of Introduction please mention how the rest of the manuscript is organized by referring to the section numbers.
Response 8
The issue has been addressed also manuscript organization is detailed, see highlighted text in red color in lines 79-82
Comment 9
In section 2.1 Define DICOM, SCLC, etc. Please note that all the abbreviations should be defined the first time they are used in the manuscript. Please also add an acronym table at the end of the manuscript.
Response 9
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 90, 385-397.
Comment 10
The link you provided www.giveascan.org looks broken. Please mention the last access date. Also, you need to put all the links in the References and treat them as regular references.
Response 10
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 87, 432.
Comment 11
Section 2.4 after “…content-based applications” please add the following articles:
https://doi.org/10.1109/SPIN.2017.8049930
https://doi.org/10.3390/s22082948
https://doi.org/10.3390/s22062346
Response 11
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 199
Comment 12
Before Equation (12), inside the {…}, please replace “a_i,” with “a_N”. Please carefully check all the equations.
Response 12
The issue has been addressed; details equations removed as suggested by reviewer 1.
Comment 13
Page 9. Please carefully check whether the values in percentage match the one in Figure 2 when you described them.
Response 13
The issue has been addressed, the whole description is carefully checked for Figure 2 and all other Figures.
Comment 14
Please write the limitations of the study and provide future direction in the conclusion part as well.
Response 14
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 367-377.
Comment 15
Please carefully check the references to ensure they are correct and have a consistent format according to the MDPI guidelines.
Response 15
The issue has been addressed the references are formatted according to Applied Sciences MDPI style
Author Response File: Author Response.docx
Round 2
Reviewer 3 Report
This work utilized different image enhancement methods to improve the lung cancer prediction using CT scan. Different machine learning based classifications are used to evaluate the classification accuracy.
Major comments:
1. The authors have stated the possible overfitting issue by adding the ‘Limitations and future recommendations’ chapter. However, without enough experiments and result analysis (average accuracy plus standard deviation), the evaluation results of the proposed method are not convincing.
2. The methods selected for comparison in Table 1 were not described. Why those methods were chosen. What are their advantage and shortage compared to the proposed method?
Author Response
Reviewer # 3:
Comment 1
- The authors have stated the possible overfitting issue by adding the ‘Limitations and future recommendations’ chapter. However, without enough experiments and result analysis (average accuracy plus standard deviation), the evaluation results of the proposed method are not convincing.
Response 1
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 94-118, 252-266, 361-401.
Comment 2
- The methods selected for comparison in Table 1 were not described. Why those methods were chosen. What are their advantage and shortage compared to the proposed method?
Response 2
The issue has been addressed and details have been incorporated, see highlighted text in red color in lines 405-422, 445-447.
Author Response File: Author Response.docx
Reviewer 4 Report
I would like to thank the authors for addressing my comments. The manuscript looks better now. I have a few more suggestions:
1. Please remove the acronym (GLCM) from the title, and use lower case letters for Extracted Gray-level Co-occurrence Matrix.
2. At the end of the Introduction, please remove "the" before "section" and instead capitalize the "S". Thus, for example, instead of "The section 1" say "Section 1", "section 2" to "Section 2", "the section 3" to "Section 3". "The section 4" to "Section 4".
Figure 6 took 2 and a half pages! Please make a figure nice so that it takes a half page or maximum one page! Please note that there are still lots of blank spaces in each panel. The legends are all the same, so just use one legend at the top of the figure horizontally for all the panels!
Similarly for Figure 3 use just one legend for the two panels and put the two panels side by side so it takes a half page maximum!
Finally please carefully proofread the manuscript before publication if accepted by the editor.
Thank you!
Author Response
Reviewer # 4:
Comment 1
- Please remove the acronym (GLCM) from the title, and use lower case letters for Extracted Gray-level Co-occurrence Matrix.
Response 1
The issue has been addressed
Comment 2
- At the end of the Introduction, please remove "the" before "section" and instead capitalize the "S". Thus, for example, instead of "The section 1" say "Section 1", "section 2" to "Section 2", "the section 3" to "Section 3". "The section 4" to "Section 4".
Response 2
The issue has been addressed
Comment 3
Figure 6 took 2 and a half pages! Please make a figure nice so that it takes a half page or maximum one page! Please note that there are still lots of blank spaces in each panel. The legends are all the same, so just use one legend at the top of the figure horizontally for all the panels!
Response 3
The issue has been addressed
Comment 4
Similarly for Figure 3 use just one legend for the two panels and put the two panels side by side so it takes a half page maximum!
Response 4
The issue has been addressed
Comment 5
Finally please carefully proofread the manuscript before publication if accepted by the editor.
Response 5
The issue has been addressed, the whole manuscript is edited for English
Author Response File: Author Response.docx
