Multi-Class Classifier in Parkinson’s Disease Using an Evolutionary Multi-Objective Optimization Algorithm
Round 1
Reviewer 1 Report
The authors have proposed a new multiple-class classification method for Parkinson’s disease MRI images. The proposed approach is based on using multi-objective evolutionary algorithm (NSGA-II) and SVM classifier. The paper is well constructed. The authors started from reviewing the related research. Next, the original solution was proposed and verified experimentally. During the experiments, a real-world dataset was used, and the obtained results were compared with those coming from other state-of-the-art approaches.
The paper is very interesting, the selected research topic is very important, and the proposed solution will find practical applications.
Please improve the English language before the publication because there are some grammar and style errors.
Author Response
Reviewer #1
Comments and Suggestions for Authors
The authors have proposed a new multiple-class classification method for Parkinson’s disease MRI images. The proposed approach is based on using multi-objective evolutionary algorithm (NSGA-II) and SVM classifier. The paper is well constructed. The authors started from reviewing the related research. Next, the original solution was proposed and verified experimentally. During the experiments, a real-world dataset was used, and the obtained results were compared with those coming from other state-of-the-art approaches.
The paper is very interesting, the selected research topic is very important, and the proposed solution will find practical applications.
Please improve the English language before the publication because there are some grammar and style errors.
Thank you so much for your kind review of our manuscript. Your comments were of great help and motivation to encourage our research. About the use of the English language, the whole paper has been comprehensively revised. An extensive track of the changes is provided below. Thanks again for your observations.
Table: Comprehensive track of changes from original to revised version
Line nr. |
Text in former contribution |
Revised version |
5 |
Each of these VOIs is subjected to volumetric feature extraction using the Discrete Wavelet Transform (3D-DWT). |
Each of these VOIs is subjected to volumetric feature extraction using the Three-Dimensional Discrete Wavelet Transform (3D-DWT). |
11 |
In order to analyze the accuracy obtained in the SVM classifier |
In order to analyze the SVM classifier accuracy |
13 |
in the multi-class classification |
in multi-class classification |
20 |
Dementia is a disease characterized by the deterioration of brain functions. |
Dementia is a clinical syndrome characterized by a progressive deterioration of cognitive functions. |
21 |
Currently, the research of brain diseases and dementia is one of the most important fields of research. |
Currently, research in brain diseases and dementia is one of the most important fields of investigation. |
28 |
due to the aging of the population |
due to the ageing of the population |
30 |
These figures show… |
|
34 |
that make it possible to slow down the development of the disease |
that slow down the development of PD |
35 |
a prolonged period |
a more extended period of time |
43 |
incoordination |
lack of coordination |
52 |
wherein rigidity inflexibility and in the legs makes the patient suffering from the disease barely have motor ability to walk \cite{Kalia2015}.. |
wherein rigidity and inflexibility in the legs makes the patient barely have motor ability to walk \cite{Kalia2015}. |
53 |
The methodology usually used in the diagnosis of PD |
The methodology usually used in PD diagnosis |
60 |
with early PD |
in early stages of PD |
73 |
MRI is a technique that has been used to study abnormal brain function in a variety of diseases. |
(Delete sentence, as it is the same idea as the following sentence). |
74 |
that has been used fruitfully |
that has been fruitfully applied |
81 |
Using the most relevant regions, an intelligent classifier, based on Support Vector Machine (SVM), is constructed the intention of being |
Using the most relevant regions, an intelligent classifier based on Support Vector Machine (SVM), is constructed with the aim of being |
84 |
(multiclass classification: Parkinson's, prodomal, SWEED and normal patients) |
(multiclass classification: Parkinson's disease, prodomal, SWEDD, gene cohort and normal patients) |
92 |
measure the initial contact to final contact of the same foot when walking and were observed as an consequence of central pattern generators established by spinal circuits |
measure the initial to final contact of the same foot when walking and were observed as a consequence of central pattern generators established by spinal circuits |
97 |
being a important |
being an important |
126 |
The MRI analysis for PD diagnosis is a very fruitful area of research, where in recent years, relevant papers have appeared developing novel methodologies |
MRI analysis for PD classification is a very fruitful area of research, indeed in recent years many relevant papers have been published developing novel methodologies |
128 |
Most of the articles sent in the bibliography focus on the realization of binary classifiers in order to determine control patients and PD patients |
Most of the articles cited in the bibliography focus on the realization of binary classifiers in order to differentiate between control and PD patients |
144 |
The data used was acquired |
Input data was acquired |
148 |
The authors |
Authors |
162 |
In recent years, Convolutional Neural Networks (CNN) has been extensively used |
In recent years, Convolutional Neural Networks (CNN) have been extensively used |
164 |
For example, in Sivaranjini et al. \cite{Sivaranjinia2020} MR images Parkinson's disease subjects and healthy control volunteers using DL neural network was used, showing an accuracy of 88.9\%, based on the PPMI data-set. |
For example, in Sivaranjini et al. \cite{Sivaranjinia2020} MR images of Parkinson's disease subjects and healthy control volunteers were classified by a DL neural network, showing an accuracy of 88.9\%, based on the PPMI data-set. |
182 |
To the best of our knowledge there are very few, if not almost no, references in the literature |
To the best of our knowledge, there are very few, if any, references in the literature |
188 |
the reduction of the dimensionality of the problem |
dimensionality reduction of the problem |
189 |
using mRMR and PCA |
applying mRMR and PCA |
205 |
(4) Use the trained SVM classifier (from training steps with selected VOI), with test samples |
(4) Classification of test samples through the SVM classifier previously trained with selected VOI |
206 |
block scheme the test phase |
block scheme of the test phase |
217 |
Phase 2: The block scheme for test phase |
Phase 2: Block scheme for the test phase |
232 |
maintaining the assumption that SWEDD patients have another motor disorder |
maintaining the assumption that SWEDD patients suffer from another motor disorder |
245 |
We must emphasize that exist patient having multiple NIfTI files |
It must be emphasized that there are patients having multiple NIfTI files |
254 |
The third group analysed |
(New paragraph) The third group which was analysed |
269 |
The first step consisted in the elimination of those duplicated or corrupted images. The final set images is reduced to a total of 917 MRI images, for the four types of patients analysed. |
The first step consisted in the elimination of duplicated and corrupted images. Afterwards, the set of images was reduced to a total of 917 MRI images, corresponding to the five types of patients that were analysed. |
270 |
SWEED group |
SWEDD group |
276 |
deviation of age and weight |
age and weight standard deviation |
279 |
Weight (Tables 1-3) |
Weight (kg) |
299 |
Although the noise that an image may have |
Although the noise that can be present in an image |
301 |
the performance of an automatic image prepossessing methodology and, therefore, it requires removal. |
the performance of an automatic image preprocessing methodology and, therefore, it requires removal. |
305 |
This features (as a real number vector) |
These features (as a real number vector) |
340 |
For example, Figure \ref{fig:wavelet} show a two level of the decomposition for an brain image. |
For example, Figure \ref{fig:wavelet} shows a two level decomposition for a brain image. |
352 |
Additionally, a better understanding of the disease by characterizing the most relevant regions of the brain for the construction of the classifier. |
Additionally, a better understanding of the disease is expected by characterizing the most relevant regions of the brain for the construction of the classifier. |
377 |
When treating a volumetric magnetic resonance image, various procedures must be performed filters |
When a volumetric magnetic resonance image is being processed, various filtering procedures must be performed |
388 |
This paper address a classification problem, in with there exist multi-class ($C$ classes) for each patterns. |
This paper addresses a classification problem where there exist multi-class ($C$ classes) for each pattern. |
400 |
any information about of $v_j$ |
any information about $v_j$ |
407 |
Determining which is the appropriate discretization is not easy task for specific problem, using in this case other |
Determining the appropriate discretization is not an easy task for this specific problem. In this case, other estimation methods of density functions as maximum likelihood, Bayesian estimation or non-parametric techniques were used. |
414 |
is not easly to be computed |
is difficult to be computed due to the fact |
416 |
In this way the insufficiency of data for the estimation of the density functions would be covered and their calculation would be would be carried out |
In this way, the lack of data for the estimation of the density functions would be solved and their calculation would be carried out |
420 |
the set of variables based on in |
the set of variables based on |
432 |
is defined as |
is defined as: |
438 |
Assuming found the suboptimal set $S_{m−1}$ of $m−1$ variables, |
Assuming the suboptimal set $S_{m−1}$ of $m−1$ variables to be found, |
439 |
to add |
to be added |
443 |
thousand |
thousands |
446 |
Indicate that |
It should be noted that |
462 |
an Multi-Objective Optimization Evolutionary |
a Multi-Objective Optimization Evolutionary |
468 |
is represented by a chromosome, and said chromosome is made up of different genes |
is represented by a chromosome, which is made up of different genes |
479 |
when said VOI is not used |
when that VOI is not used |
490 |
The population size used in the simulation carried out was 250. |
The population size that was used in the simulation was 250. |
495 |
that have their value to one |
that are set to one |
496 |
elist strategy: |
elitist strategy: |
513 |
typically the maximum number of generations exceeded |
typically, a maximum number of generations is reached |
518 |
(measured as the number of VOIs used ) |
(measured as the number of needed VOIs). |
524 |
like the one presented in the figure |
like the one presented in Figure \ref{fig:Solution_ParetoFront} |
530 |
a feature reduction is performed |
a feature reduction procedure is performed |
553 |
and test phase of the SVM classifier obtained |
and a test phase of the obtained SVM classifier |
558 |
potential solutions, called chromosome |
i.e. potential solutions, called chromosomes |
564 |
execution times |
execution time |
575 |
In multi-class problems is frequently used the confusion matrix |
In multi-class problems the confusion matrix is frequently used |
591 |
the precision decreases |
precision decreases |
618 |
As was presented |
As it was described |
632 |
shows the location |
show the location |
637 |
It is relevant to make a comparison of the results in classification of PD disease |
It is relevant to make a comparison of the PD disease classification results |
642 |
Regions of the brain selected (Fig. 14-15) |
Selected brain regions |
644 |
The novel methodology presented |
The novel presented methodology |
645 |
classification of Parkinson's disease is of relevance to expert neurologists |
classification of Parkinson's disease is of great relevance to expert neurologists |
647 |
Obtaining the most relevant VOI was thanks to the implementation |
Obtaining the most relevant VOIs was accomplished due to the implementation |
677 |
The results obtained in the confusion matrices, for the different solutions of the MOE analyzed, |
Results from the confusion matrices for the different solutions of the analyzed MOE |
Author Response File: Author Response.pdf
Reviewer 2 Report
This paper describes a novel methodology to identify patients with Parkinson's disease and other related cases. What is novel about this study is the use of a multiclass approach combined with multi-objective evolutionary computation in order to face the problem.
For me, this paper presents a complete state of the art, an appropriate research methodology, a clear description of the results and a good conclusion supported by them.
I recommend accepting the work, as long as the authors can attend to some minor details that I have detected in the writing:
1) Page 1. Line 30. "These figures show..." I don't understand why authors use the word "figures" at the beginning of this sentence. what figures?
2) PAge 2. Line 73 & 74. The two sentences express the same idea. Remove one of them.
3) Page 4. Line 145. It is necessary to report the accuracy of the work that is being cited.
4) Tables 1, 2, and 3. It can be inferred that the units of the reported age is in years. but what about the weight? which units were used at these tables?
5)There are several typos, duplicate words, and misspelling errors throughout the work. Authors are advised to do a careful review. Below are some of them
a) page 2. Line 51. Remove double period.
b) page 2. Line 72 Remove "is a that".
c) Page 3. Line 128. Add the word "Table" to reference to Table 7
d) Page 10. Line 364. Remove period between "Ny " and "Nz", and put z as a subscript
e) Page 12. Line 416. Remove duplicated "would be"
Author Response
Reviewer #2
Comments and Suggestions for Authors
This paper describes a novel methodology to identify patients with Parkinson's disease and other related cases. What is novel about this study is the use of a multiclass approach combined with multi-objective evolutionary computation in order to face the problem.
For me, this paper presents a complete state of the art, an appropriate research methodology, a clear description of the results and a good conclusion supported by them.
Thank you for your knowledgeable revision of this manuscript. Your comments were of great help to improve our contribution. Please, let us let you know that the whole paper has been comprehensively revised, so that English language errors have been minimized. An extensive track of the changes has been provided at the end of this reply.
I recommend accepting the work, as long as the authors can attend to some minor details that I have detected in the writing:
1) Page 1. Line 30. "These figures show..." I don't understand why authors use the word "figures" at the beginning of this sentence. what figures?
Sentence on line 30 was changed to “These evidences show the urgent need to advance in Parkinson's research, since an early diagnosis helps to delay the development of the disease in the patient.”.
2) PAge 2. Line 73 & 74. The two sentences express the same idea. Remove one of them.
Thank you for your suggestion. Indeed, the sentence “MRI is a technique that has been used to study abnormal brain function in a variety of diseases.” was eliminated has it is redundant with the next sentence.
3) Page 4. Line 145. It is necessary to report the accuracy of the work that is being cited.
Thank you for your suggestion. Indeed, the reviewer is right and although it had been commented in table 7 of this contribution a resume of the accuracy of the different methods presented in the bibliography, it is necessary to indicate it in this section the accuracy of the contribution of Lei et al. This data has been included at the end of the paragraph:
“In Lei et al. \cite{Lei2018}, a multi-class classification system for PD analysis, based on a sparse discriminative feature selection methodology, is presented. The authors presented a framework to create a least square regression model based on Fisher's linear discriminant analysis and locality preserving projection. Input data was acquired from the Parkinson's Progression Markers Initiative (PPMI) database, using 123 PD, 56 normal and 29 SWEDD images, obtaining an accuracy of 78.4\%. “
4) Tables 1, 2, and 3. It can be inferred that the units of the reported age is in years. but what about the weight? which units were used at these tables?
Thank you for the advice, the used weight unit along the paper is the kilogram, and it was accordingly specified in tables “weight (kg)”.
5)There are several typos, duplicate words, and misspelling errors throughout the work. Authors are advised to do a careful review. Below are some of them
- a) page 2. Line 51. Remove double period.
- b) page 2. Line 72 Remove "is a that".
- c) Page 3. Line 128. Add the word "Table" to reference to Table 7
- d) Page 10. Line 364. Remove period between "Ny " and "Nz", and put z as a subscript
- e) Page 12. Line 416. Remove duplicated "would be"
All of these misprints and errors were fixed as well as careful review was done along the whole manuscript. Thanks again for your corrections.
Table: Comprehensive track of changes from original to revised version
Line nr. |
Text in former contribution |
Revised version |
5 |
Each of these VOIs is subjected to volumetric feature extraction using the Discrete Wavelet Transform (3D-DWT). |
Each of these VOIs is subjected to volumetric feature extraction using the Three-Dimensional Discrete Wavelet Transform (3D-DWT). |
11 |
In order to analyze the accuracy obtained in the SVM classifier |
In order to analyze the SVM classifier accuracy |
13 |
in the multi-class classification |
in multi-class classification |
20 |
Dementia is a disease characterized by the deterioration of brain functions. |
Dementia is a clinical syndrome characterized by a progressive deterioration of cognitive functions. |
21 |
Currently, the research of brain diseases and dementia is one of the most important fields of research. |
Currently, research in brain diseases and dementia is one of the most important fields of investigation. |
28 |
due to the aging of the population |
due to the ageing of the population |
30 |
These figures show… |
|
34 |
that make it possible to slow down the development of the disease |
that slow down the development of PD |
35 |
a prolonged period |
a more extended period of time |
43 |
incoordination |
lack of coordination |
52 |
wherein rigidity inflexibility and in the legs makes the patient suffering from the disease barely have motor ability to walk \cite{Kalia2015}.. |
wherein rigidity and inflexibility in the legs makes the patient barely have motor ability to walk \cite{Kalia2015}. |
53 |
The methodology usually used in the diagnosis of PD |
The methodology usually used in PD diagnosis |
60 |
with early PD |
in early stages of PD |
73 |
MRI is a technique that has been used to study abnormal brain function in a variety of diseases. |
(Delete sentence, as it is the same idea as the following sentence). |
74 |
that has been used fruitfully |
that has been fruitfully applied |
81 |
Using the most relevant regions, an intelligent classifier, based on Support Vector Machine (SVM), is constructed the intention of being |
Using the most relevant regions, an intelligent classifier based on Support Vector Machine (SVM), is constructed with the aim of being |
84 |
(multiclass classification: Parkinson's, prodomal, SWEED and normal patients) |
(multiclass classification: Parkinson's disease, prodomal, SWEDD, gene cohort and normal patients) |
92 |
measure the initial contact to final contact of the same foot when walking and were observed as an consequence of central pattern generators established by spinal circuits |
measure the initial to final contact of the same foot when walking and were observed as a consequence of central pattern generators established by spinal circuits |
97 |
being a important |
being an important |
126 |
The MRI analysis for PD diagnosis is a very fruitful area of research, where in recent years, relevant papers have appeared developing novel methodologies |
MRI analysis for PD classification is a very fruitful area of research, indeed in recent years many relevant papers have been published developing novel methodologies |
128 |
Most of the articles sent in the bibliography focus on the realization of binary classifiers in order to determine control patients and PD patients |
Most of the articles cited in the bibliography focus on the realization of binary classifiers in order to differentiate between control and PD patients |
144 |
The data used was acquired |
Input data was acquired |
148 |
The authors |
Authors |
162 |
In recent years, Convolutional Neural Networks (CNN) has been extensively used |
In recent years, Convolutional Neural Networks (CNN) have been extensively used |
164 |
For example, in Sivaranjini et al. \cite{Sivaranjinia2020} MR images Parkinson's disease subjects and healthy control volunteers using DL neural network was used, showing an accuracy of 88.9\%, based on the PPMI data-set. |
For example, in Sivaranjini et al. \cite{Sivaranjinia2020} MR images of Parkinson's disease subjects and healthy control volunteers were classified by a DL neural network, showing an accuracy of 88.9\%, based on the PPMI data-set. |
182 |
To the best of our knowledge there are very few, if not almost no, references in the literature |
To the best of our knowledge, there are very few, if any, references in the literature |
188 |
the reduction of the dimensionality of the problem |
dimensionality reduction of the problem |
189 |
using mRMR and PCA |
applying mRMR and PCA |
205 |
(4) Use the trained SVM classifier (from training steps with selected VOI), with test samples |
(4) Classification of test samples through the SVM classifier previously trained with selected VOI |
206 |
block scheme the test phase |
block scheme of the test phase |
217 |
Phase 2: The block scheme for test phase |
Phase 2: Block scheme for the test phase |
232 |
maintaining the assumption that SWEDD patients have another motor disorder |
maintaining the assumption that SWEDD patients suffer from another motor disorder |
245 |
We must emphasize that exist patient having multiple NIfTI files |
It must be emphasized that there are patients having multiple NIfTI files |
254 |
The third group analysed |
(New paragraph) The third group which was analysed |
269 |
The first step consisted in the elimination of those duplicated or corrupted images. The final set images is reduced to a total of 917 MRI images, for the four types of patients analysed. |
The first step consisted in the elimination of duplicated and corrupted images. Afterwards, the set of images was reduced to a total of 917 MRI images, corresponding to the five types of patients that were analysed. |
270 |
SWEED group |
SWEDD group |
276 |
deviation of age and weight |
age and weight standard deviation |
279 |
Weight (Tables 1-3) |
Weight (kg) |
299 |
Although the noise that an image may have |
Although the noise that can be present in an image |
301 |
the performance of an automatic image prepossessing methodology and, therefore, it requires removal. |
the performance of an automatic image preprocessing methodology and, therefore, it requires removal. |
305 |
This features (as a real number vector) |
These features (as a real number vector) |
340 |
For example, Figure \ref{fig:wavelet} show a two level of the decomposition for an brain image. |
For example, Figure \ref{fig:wavelet} shows a two level decomposition for a brain image. |
352 |
Additionally, a better understanding of the disease by characterizing the most relevant regions of the brain for the construction of the classifier. |
Additionally, a better understanding of the disease is expected by characterizing the most relevant regions of the brain for the construction of the classifier. |
377 |
When treating a volumetric magnetic resonance image, various procedures must be performed filters |
When a volumetric magnetic resonance image is being processed, various filtering procedures must be performed |
388 |
This paper address a classification problem, in with there exist multi-class ($C$ classes) for each patterns. |
This paper addresses a classification problem where there exist multi-class ($C$ classes) for each pattern. |
400 |
any information about of $v_j$ |
any information about $v_j$ |
407 |
Determining which is the appropriate discretization is not easy task for specific problem, using in this case other |
Determining the appropriate discretization is not an easy task for this specific problem. In this case, other estimation methods of density functions as maximum likelihood, Bayesian estimation or non-parametric techniques were used. |
414 |
is not easly to be computed |
is difficult to be computed due to the fact |
416 |
In this way the insufficiency of data for the estimation of the density functions would be covered and their calculation would be would be carried out |
In this way, the lack of data for the estimation of the density functions would be solved and their calculation would be carried out |
420 |
the set of variables based on in |
the set of variables based on |
432 |
is defined as |
is defined as: |
438 |
Assuming found the suboptimal set $S_{m−1}$ of $m−1$ variables, |
Assuming the suboptimal set $S_{m−1}$ of $m−1$ variables to be found, |
439 |
to add |
to be added |
443 |
thousand |
thousands |
446 |
Indicate that |
It should be noted that |
462 |
an Multi-Objective Optimization Evolutionary |
a Multi-Objective Optimization Evolutionary |
468 |
is represented by a chromosome, and said chromosome is made up of different genes |
is represented by a chromosome, which is made up of different genes |
479 |
when said VOI is not used |
when that VOI is not used |
490 |
The population size used in the simulation carried out was 250. |
The population size that was used in the simulation was 250. |
495 |
that have their value to one |
that are set to one |
496 |
elist strategy: |
elitist strategy: |
513 |
typically the maximum number of generations exceeded |
typically, a maximum number of generations is reached |
518 |
(measured as the number of VOIs used ) |
(measured as the number of needed VOIs). |
524 |
like the one presented in the figure |
like the one presented in Figure \ref{fig:Solution_ParetoFront} |
530 |
a feature reduction is performed |
a feature reduction procedure is performed |
553 |
and test phase of the SVM classifier obtained |
and a test phase of the obtained SVM classifier |
558 |
potential solutions, called chromosome |
i.e. potential solutions, called chromosomes |
564 |
execution times |
execution time |
575 |
In multi-class problems is frequently used the confusion matrix |
In multi-class problems the confusion matrix is frequently used |
591 |
the precision decreases |
precision decreases |
618 |
As was presented |
As it was described |
632 |
shows the location |
show the location |
637 |
It is relevant to make a comparison of the results in classification of PD disease |
It is relevant to make a comparison of the PD disease classification results |
642 |
Regions of the brain selected (Fig. 14-15) |
Selected brain regions |
644 |
The novel methodology presented |
The novel presented methodology |
645 |
classification of Parkinson's disease is of relevance to expert neurologists |
classification of Parkinson's disease is of great relevance to expert neurologists |
647 |
Obtaining the most relevant VOI was thanks to the implementation |
Obtaining the most relevant VOIs was accomplished due to the implementation |
677 |
The results obtained in the confusion matrices, for the different solutions of the MOE analyzed, |
Results from the confusion matrices for the different solutions of the analyzed MOE |
Author Response File: Author Response.pdf