Deep-Learning-Based Stress-Ratio Prediction Model Using Virtual Reality with Electroencephalography Data

Round 1

Reviewer 1 Report

The aim of this study was to present a stress ratio prediction model by monitoring the building space information in electroencephalography (EEG) information. Authors create very interesting VR environment and measure some indicators of stress and tension manifesting mainly in the EEG beta waves. In future, in my opinion, autohors should take into consideration minimizing the effect of various artifacts originating from moving participants and interaction of the EEG and VR equipment. Even though, there are many papers about stress and EEG, fMRI, etc. the approach of authors make this paper worty publishing. 

Author Response

Point 1: The aim of this study was to present a stress ratio prediction model by monitoring the building space information in electroencephalography (EEG) information. Authors create very interesting VR environment and measure some indicators of stress and tension manifesting mainly in the EEG beta waves. In future, in my opinion, autohors should take into consideration minimizing the effect of various artifacts originating from moving participants and interaction of the EEG and VR equipment. Even though, there are many papers about stress and EEG, fMRI, etc. the approach of authors make this paper worty publishing. 

 

Response 1: Thanks for the review.

Reviewer 2 Report

 

 

 

 Deep Learning-based Stress Ratio Prediction Model  Using Virtual Reality with Electroencephalography  Data

 

The paper presents a very interesting topic that has wide applications in modern societies. However, there are some major points that should be addressed to improve the presentation of the manuscript.

 

• The abstract should be update such as to include the finding of this work and the conclusion or recommendation.
• In section 2.1, the characteristics of EEG bands are well established and should be removed (from line 74 to line 99, should be removed).
• In the theoretical consideration section 2.1. The author should provide more literature on EEG and Stress. For example, stress level has been assessed using prefrontal EEG signals/fNIRS and machine learning approach in many previous studies including but not limited to:
• Al-Shargie, F. M., Tong Boon Tang, Nasreen Badruddin, and Masashi Kiguchi. "Mental stress quantification using EEG signals." In International Conference for Innovation in Biomedical Engineering and Life Sciences, pp. 15-19. Springer, Singapore, 2015.
• Al-Shargie, Fares, Tong Boon Tang, Nasreen Badruddin, and Masashi Kiguchi. "Towards multilevel mental stress assessment using SVM with ECOC: an EEG approach." Medical & biological engineering & computing56, no. 1 (2018): 125-136.
• Alyan, Emad, Naufal M. Saad, and Nidal Kamel. "Effects of Workstation Type on Mental Stress: fNIRS Study." Human Factors(2020): 0018720820913173.
• Al-Shargie, Fares, Tong Boon Tang, and Masashi Kiguchi. "Assessment of mental stress effects on prefrontal cortical activities using canonical correlation analysis: an fNIRS-EEG study." Biomedical optics express8, no. 5 (2017): 2583-2598.
• Al-Shargie, Fares, Masashi Kiguchi, Nasreen Badruddin, Sarat C. Dass, Ahmad Fadzil Mohammad Hani, and Tong Boon Tang. "Mental stress assessment using simultaneous measurement of EEG and fNIRS." Biomedical optics express7, no. 10 (2016): 3882-3898.

 

• The data acquisition time and sampling frequency should be add to the experimental measurement and method section.
• The number of participants and the time of the experiment are missing.
• The virtual spaces were not randomized.
• The data preprocessing is missing.
• It is not clear how the EEG frequency bands were extracted from the raw data.
• In figure 9 it is not clear what is x-axis and y-axis stands for.
• The study is lack of discussion with the state of the art EEG stress studies.
• The effects of circadian rhythm should be discussed in this study.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The objective of the paper is an experimental study to measure the amount of stress change from electroencephalographic (EEG) signals from subjects while visiting a virtual space of a particular historical building. Results show correlation between emotional state and some places in the building by using a stress index and band frequency analysis of the EEG signals. The methods employed are well-known and there is no theoretical contribution provided. However, experimental results could be interesting for practical purposes. Literal presentation of the paper has room for improvement, there are some typos, misused terms, and readability of figures could be enhanced. Temporal dependency of the dynamic of brain oscillations could be related with the states experimented by subjects. This should be discussed in the paper. Besides, some aspects of the data analysis should be improved including giving details of variance and statistical significance of the experimental results. In summary, I consider the contents of the paper are potentially publishable, but the following specific issues should be addressed in a revised version of the paper.

- Literal presentation of the paper has much room for improvement. There are several inaccuracies that have to be corrected. For instance: (i) The term “artificial intelligence (AI)” should be changed. AI is a global term that covers many methodologies that are not employed in the paper. More suitable terms for the particular method used in the paper could be “machine learning”, “pattern recognition”, “automatic classification”, or directly use the name of the classifier. (ii) Lines 21, 60, please explain what you mean by 10 bits? (iii) Lines 108, 409, please add a rationale as to why this application is big data. (iv) Line 64, “a platform”, “a virtual platform”? (v) Lines 108-109, “collected at 60 or more data points per second.”, it seems various sampling frequencies were used, why?; besides, I would include the acronym “Hz” in that sentence. (vi) Line 119, “It is one of” instead of “It one of”. (vii) Line 187, “SGD” is not defined. All the acronyms used in the paper should be defined in the text of the paper the first time they appear, except in the abstract. (viii) Line 192, “Figure 3” instead of “Figure 2”? (ix) Line 201, “K University”, please use full name of the university; (x) Line 222, “Additionally, it is cheaper than other devices on the market”, this sentence is irrelevant.

Therefore, a comprehensive English proof-reading of the paper should be done.

- Several figures should be modified to improve readability. For instance, (i) font sizes should be enlarged in Figure 1; (ii) Figure 2, clarify what is before and after demolition; (iii) Figure 4, change background to white and line of the gray curves to black to improve readability of curves, enlarge font size; (iv) Figure 8, enlarge font sizes, y axis title horizontally aligned, show units of y axis or comment that in figure caption; (v) Figure 9, label y axis, meaning of colors, decrease the number of x tick labels.

- Other case studies could be mentioned in Introduction such as Stendhal syndrome that relates emotional states to observing artworks.

- More bibliographic references are needed to support the ideas in the paragraph from line 145 to 161. In addition, please include a reference to “Tableau” (line 298).

- Temporal dependency of the dynamic of brain oscillations has been shown to be related with the states experimented by subjects using EEG signal analysis. The EEG data from the described experiment in the paper correspond to a long term non-stationary process with nonlinearities and temporal dependence. This should be considered for implementation of competitive analysis methods for the paper experiment such as sequential independent component analysis models (SICAMM) and long short-term memory (LSTM) recurrent neural networks. Please discuss about this. I suggest the following references: [1] “On including sequential dependence in ICA mixture models,” Signal Processing 90 (7), pp. 2314-2318, 2010. [2] “Multichannel dynamic modeling of non-Gaussian mixtures,” Pattern Recognition 93, pp. 312-323, 2019.

- Please provide the following data of the experiment: (i) duration of the different stages, including preparation of the subjects and preprocessing of the signals. (ii) I wonder data analysis is off-line, please specify it. Besides, discuss about real-time implementation of the proposed application.

- The following aspects of the data analysis should be improved: (i) the number and characteristics (age, sex…) of subjects should be specified; (ii) the mean and standard deviation of all experiments should be shown and analyzed. The validation of statistical significance evaluation of the results should be improved by implementing a proper statistical test to evaluate the probability with which one can say that the two distributions are significantly different, or whether the difference in the mean is just caused by statistical fluctuations.

 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The author has addressed all my comments. However the content can be furthered improved. Please correct the following: 

1- In Line 111 and 112, the references style were not consistent. Please use either a number i.e [1] or the name of author and year in all the references. Please follow the guideline of the journal.

2- in line 382, the Figure 9 (A) is not correct , it should be Figure 10 (A).

3- The statistical analysis of t-test should be further improved so that the reader can easily understand it. for example what did you compare in the analysis (X,Y) what is X and what is Y.

4- Why only 1 subject was recruited in this study? This is the main weakness of this paper. 

Author Response

Point 1:  In Line 111 and 112, the references style were not consistent. Please use either a number i.e [1] or the name of author and year in all the references. Please follow the guideline of the journal.

Response 1: Changed the reference style to number [1].

 

Point 2: in line 382, the Figure 9 (A) is not correct , it should be Figure 10 (A).

Response 2: Figure 9 (A) has been changed to Figure 10 (A).

 

Point 3: The statistical analysis of t-test should be further improved so that the reader can easily understand it. for example what did you compare in the analysis (X,Y) what is X and what is Y.

Response 3: A t-test was performed to analyse the information of the net stress in the first(Beta Waves Only) and secondary(alpha, beta, theta, gamma, and delta waves) analyses.

 

Point 4: Why only 1 subject was recruited in this study? This is the main weakness of this paper. 

Response 4: The main purpose of this paper is to analyze the architectural space.

We examined changes in brainwaves according to the variation in the pattern of building spaces, in which the variables of brainwaves were analysed per frequency using a deep-learning algorithm to find a section with a high level of stress. Therefore, the number of subjects was set to 1 for the experiment.

 

Thank you for the review.

 

 

Reviewer 3 Report

The quality of the paper has been improved significantly. All my concerns have been adequately addressed in the revised version of the paper. The authors have made a great effort to improve the quality of literal presentation and figures of the paper; the details and evaluation of experimentation; and the related bibliography. I already considered the paper to be acceptable, except for very minor details, and I confirm my judgement. In summary, I think the paper should be ready for publication after final editing revision. Minor changes: please standardize the format of bibliographic references.

Author Response

Point 1: Minor changes: please standardize the format of bibliographic references.

 

Response 1: Revised references. Thank you for the review.