Melanoma Detection in Dermoscopic Images Using a Cellular Automata Classifier
, José Cruz Martínez-Perales 1, Jorge Cortés-Galicia 1, Rolando Flores-Carapia 2 and Víctor Manuel Silva-García 2Round 1
Reviewer 1 Report
The paper proposes a method to detect melanoma in dermatoscopic images.
This is well-written but there are some typos.
On page 2, “using convolutional networks” should be “using convolutional neural networks.”
On page 2, “Alex Net” should be “AlexNet.”
On page 2, “Vector Support Machine” should be “Support Vector Machine.”
On page 7, “Standar” should be “Standard.”
Also, in Table 1, what does it mean by “Etp - …”? Isn’t is “Etp = …”? Is the Modified Entropy formula correct? Doesn’t it need a minus (-) sign? Please double check it.
Author Response
Good afternoon dear Reviewer. First of all, on behalf of those who make up the work team, we want to thank you for your valuable comments to improve the work presented. Responses to your comments are shown below.
Comments and Suggestions for Authors
The paper proposes a method to detect melanoma in dermatoscopic images.
This is well-written but there are some typos.
On page 2, “using convolutional networks” should be “using convolutional neural networks.”
Answer: The observation has been corrected. Page 2, penultimate line of subsection 1.2 Related jobs.
On page 2, “Alex Net” should be “AlexNet.”
Answer: the observation was corrected. Page 2. Penultimate line of subsection 1.2. Related jobs.
On page 2, “Vector Support Machine” should be “Support Vector Machine.”
Answer: the observation was corrected. Page 2. Penultimate line of subsection 1.2. Related jobs.
On page 7, “Standar” should be “Standard.”
Answer: the observation was corrected. Page 7, table 1.
Also, in Table 1, what does it mean by “Etp - …”? Isn’t is “Etp = …”? Is the Modified Entropy formula correct? Doesn’t it need a minus (-) sign? Please double check it.
Answer: The symbol "=" was needed in the definition of entropy (Etp). The observation was corrected. Page 7, table 1. On the other hand, the modified entropy formula is correct, can be seen in reference [25]. To the formulas in table 1 I added (those that were missing) the notation of the term that they define, for example, for Skewness (Sk), for Kurtosis (K), for Uniformity (U) and for Modified Entropy (MEtp ).
Reviewer 2 Report
- 2nd sentence of introduction: "The non-melanoma type ... of the body [3,4]." --> Basal cell and squamous cell carcinomas are non-melanoma malignancies that occur in the epidermis. They present as nodules or nonpainful ulcerated and crusted lesions that do not heal with the passage of time. Their growth is slow, and they rarely metastasize [3,4].
- last sentence of the 1st paragraph of the introduction: "Diagnosis of melanoma ... reproducibility." --> Diagnosis of melanoma is a challenging task due to the similarity in appearance to nonmalignant nevi.
- Acknowledgments: "economical" --> financial
- Legend to table 3: Define ACC, SE & SP here so that the table can stand on its own.
- reference 38: "clas-specific" --> class-specific
Author Response
Good afternoon dear Reviewer. First of all, on behalf of those who make up the work team, we want to thank you for your valuable comments to improve the work presented. Responses to your comments are shown below.
Comments and Suggestions for Authors
- 2nd sentence of introduction: "The non-melanoma type ... of the body [3,4]." --> Basal cell and squamous cell carcinomas are non-melanoma malignancies that occur in the epidermis. They present as nodules or nonpainful ulcerated and crusted lesions that do not heal with the passage of time. Their growth is slow, and they rarely metastasize [3,4].
Answer: Your comment was considered and the text was changed as suggested.
- last sentence of the 1st paragraph of the introduction: "Diagnosis of melanoma ... reproducibility." --> Diagnosis of melanoma is a challenging task due to the similarity in appearance to nonmalignant nevi.
Answer: Your comment was considered and the text was changed as suggested.
- Acknowledgments: "economical" --> financial
Answer: Your comment was considered and the text was changed as suggested.
- Legend to table 3: Define ACC, SE & SP here so that the table can stand on its own.
Answer: The terms ACC, SE and SP were defined in the legend of Table 3.
- reference 38: "clas-specific" --> class-specific
Answer: Corrected the text of the observation.
Reviewer 3 Report
"Detection of Melanoma in Dermatoscopic Images using a Classifier based on Cellular Automata"
It is interesting to evaluate the performance of the melanoma detection model in dermoscopy images using the cellular automaton-based classifier you proposed. However, there are a few corrections that are essential to meet the standard for publication. Please refer to the following comments.
1) The performance of your model is very good. However, there was a slight misclassification. There is no consideration of their causes or remedies. Please add false positives, false negative ratings and considerations.
2) With the development of deep learning in recent years, classification by many deep learning models has also been reported. Please consider the advantages and disadvantages of your model and deep learning models, other than accuracy.
Is your model practical for use in clinical medicine?
Author Response
Good afternoon dear Reviewer. First of all, on behalf of those who make up the work team, we want to thank you for your valuable comments to improve the work presented. Responses to your comments are shown below.
Comments and Suggestions for Authors
"Detection of Melanoma in Dermatoscopic Images using a Classifier based on Cellular Automata"
It is interesting to evaluate the performance of the melanoma detection model in dermoscopy images using the cellular automata-based classifier you proposed. However, there are a few corrections that are essential to meet the standard for publication. Please refer to the following comments.
1) The performance of your model is very good. However, there was a slight misclassification. There is no consideration of their causes or remedies. Please add false positives, false negative ratings and considerations.
Answer: Table 2 shows the confusion matrix where the registered values of correctly classified data are obtained, as well as that of false positives and false negatives. However, to emphasize these data, the false positive and false negative data and considerations of the possible reasons why the model erroneously detected them were added in the discussion part. See section 5 (discussion), line 1.
2) With the development of deep learning in recent years, classification by many deep learning models has also been reported. Please consider the advantages and disadvantages of your model and deep learning models, other than accuracy.
Answer: Yes, the development of deep learning has had a huge boom, however, there are other approaches to learning, mainly supervised, which is what we focus on, where there are a variety of competitive algorithms. When proposing a classification algorithm using cellular automata, the main advantages are that they are easy to implement and with low computational resources compared to deep learning. However, due to this great boom in the development of deep learning there are architectures that carry the classification with a high degree of accuracy, however, by the "No Free Lunch" theorem it establishes that, given a couple of search algorithms, there will be so many problems for which the first algorithm is better than the second, such as problems for which the second algorithm will be better than the first, concluding that there is no algorithm that is universally better than the others. By proposing this algorithm for classification by cellular automata, what we propose is to apply concepts that exist in the field of cellular automata and that could be brought to the field of supervised learning through this approach, that of cellular automata. Added paragraph at the end of subsection 1.2.
Is your model practical for use in clinical medicine?
Answer: The proposed model is applied to the detection of melanoma, and for this, the PH2 image bank obtained in [36] was used, which was developed for research purposes in order to facilitate comparative studies on segmentation and classification algorithms of dermoscopic images. These images were acquired at the Dermatology Service of the Pedro Hispano Hospital, Matosinhos, Portugal. To bring it to clinical use, it would be necessary to feed it with more images that the experts in the field of medicine obtain, in addition to possibly adapting the images to the proposed model, since, although dermoscopy studies make use of a dermatoscope, for the acquisition of the image is usually used a photographic sensor and from here it depends on the quality of the images acquired.
Reviewer 4 Report
Abstract
The abstract is not well developed. The background is too long, while, materials and methods and results must to be improved.
Computer-aided diagnostic (CAD)
CAD is Computer Aided Design. Please change it accordingly.
Introduction
Computer-assisted systems (CAD)
CAD is Computer Aided Design. Please change it accordingly.
Related jobs
Please, consider to split this section. In the introduction section I would like to read just a background, introducing the concepts and their limitation, and why authors propose a new research. On the other hands, comparison of the proposed method with other methods must to be reported in the discussion sections.
The study design must to be reported in the title, in the abstract, at the and of the introduction (The aim of the present …) in the materials and methods (This study was designed as …), in the discussion, together with the limitations (The main limitations of the present …).
Materials and methods section is missed. Please check the journal guidelines.
Experiments and results.
How the images were selected must to be described. Overall sample, inclusion and exclusion criteria, how selected the image, etc. This is a crucial point. Later, primary and secondary outcomes, statistical analysis , etc. Results section must to be developed accordingly.
Materials and methods, and results should be separate sections. Please check journal guidelines.
Discussion section must to be improved. In my opinion, most of the concept in the introduction section could be moved in the discussion
Conclusion must to be shortened. Similarly, most of the statements reported in the conclusion must to be moved in the discussion. Conclusion must to report just the conclusion supported by the results.
Did the study approved by an ethical committee?
Author Response
Good afternoon dear Reviewer. First of all, on behalf of those who make up the work team, we want to thank you for your valuable comments to improve the work presented. Responses to your comments are shown below.
Comments and Suggestions for Authors
Abstract
The abstract is not well developed. The background is too long, while, materials and methods and results must to be improved.
Answer: The Experiments and Results, Discussion and Conclusions sections have been improved. On the other hand, the summary shows the concepts from cancer to the results obtained in a summarized way, and thus, the information stands on its own, that is, the information is self-contained.
Computer-aided diagnostic (CAD)
CAD is Computer Aided Design. Please change it accordingly.
Answer: The text of the observation in the abstract was corrected.
Introduction
Computer-assisted systems (CAD)
CAD is Computer Aided Design. Please change it accordingly.
Answer: Corrected comment text in subsection 1.2.
Related jobs
Please, consider to split this section. In the introduction section I would like to read just a background, introducing the concepts and their limitation, and why authors propose a new research. On the other hands, comparison of the proposed method with other methods must to be reported in the discussion sections.
Answer: A paragraph was added at the end of subsection 1.2 where the background of the proposed work is exposed. The reasons why we propose a classifier using cellular automata is that as a research team we are working on making use of concepts that exist in the field of cellular automata and taking them to the field of pattern recognition in order to obtain improvements in this first proposed model. In addition, we want to apply not only the concepts to the classifier itself, but to the images themselves, that is, instead of making use of convolutions on the images, apply a cellular automata, however, this is a future work that we have. And this is also because, by their nature, cellular automata are easy to implement and use few computational resources compared to other models. In conclusion, we wish to create a bridge between the field of cellular automata and that of pattern recognition.
The study design must to be reported in the title, in the abstract, at the and of the introduction (The aim of the present …) in the materials and methods (This study was designed as …), in the discussion, together with the limitations (The main limitations of the present …).
Answer: Recommendations were considered in the introduction (last paragraph of the introduction) and in the discussion section (last paragraph of the discussion section).
Materials and methods section is missed. Please check the journal guidelines.
Answer: The legend of point 3. The model proposed by 3. Materials and methods has been changed.
Experiments and results.
How the images were selected must to be described. Overall sample, inclusion and exclusion criteria, how selected the image, etc. This is a crucial point. Later, primary and secondary outcomes, statistical analysis , etc. Results section must to be developed accordingly.
Answer: A description of how the images were selected was added, in fact only 10 of the total that PH2 database has were discarded, and the document describes the reasons why they were discarded.
Materials and methods, and results should be separate sections. Please check journal guidelines.
Answer: Materials and methods are shown in section 3, and experiments and results are shown in section 4.
Discussion section must to be improved. In my opinion, most of the concept in the introduction section could be moved in the discussion
Answer: The discussion section was improved by moving part of the conclusions as proposed to the discussion section, also the part referring to false positives, false negatives and considerations was added.
Conclusion must to be shortened. Similarly, most of the statements reported in the conclusion must to be moved in the discussion. Conclusion must to report just the conclusion supported by the results.
Answer: The conclusions section was shortened (reporting the conclusions supported by the results) and part of it moved to the discussion section as recommended.
Did the study approved by an ethical committee?
Answer: The set of dermoscopic images obtained in PH2 database [36] was developed for research purposes in order to facilitate comparative studies on algorithms for segmentation and classification of dermoscopic images. These images were acquired at the Dermatology Service of Hospital Pedro Hispano, Matosinhos, Portugal, and it was supported by the Portuguese Foundation for Science and Technology (FCT), so we do not know if the images considered were approved by an ethics committee. On the other hand, the work that we propose using this set of images was not approved by an ethics committee.
Round 2
Reviewer 4 Report
Dear authors, I am fine with almost all the revisions. I still suggest to shortened the conclusions. The first 6 lines are a repetition of the text. Just report conclusions and further evaluations needed. English language must to be improved.
Author Response
Good afternoon dear Reviewer. First of all, on behalf of those who make up the work team, we want to thank you for your valuable comments to improve the work presented. Responses to your comments are shown below.
Comments and Suggestions for Authors
Dear authors, I am fine with almost all the revisions. I still suggest to shortened the conclusions. The first 6 lines are a repetition of the text. Just report conclusions and further evaluations needed. English language must to be improved.
Answer: Your observation was considered as suggested by reporting only the conclusions in the respective section (section 6. Conclusions). A thorough review of the English language was conducted.
