Next Article in Journal
Trehalose: Neuroprotective Effects and Mechanisms—An Updated Review
Previous Article in Journal
Discontinuation of Cerebro-Spinal Fluid (CSF) Drainage in Acute Hydrocephalus: A Prospective Cohort Study and Exploratory Data Analysis
 
 
Article
Peer-Review Record

Wearable EEG-Based Brain–Computer Interface for Stress Monitoring

NeuroSci 2024, 5(4), 407-428; https://doi.org/10.3390/neurosci5040031
by Brian Premchand 1, Liyuan Liang 1, Kok Soon Phua 1, Zhuo Zhang 1, Chuanchu Wang 1, Ling Guo 1, Jennifer Ang 2, Juliana Koh 2, Xueyi Yong 2 and Kai Keng Ang 1,3,*
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
NeuroSci 2024, 5(4), 407-428; https://doi.org/10.3390/neurosci5040031
Submission received: 21 August 2024 / Revised: 25 September 2024 / Accepted: 25 September 2024 / Published: 8 October 2024

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this work, the authors propose a portable system that measures EEG signals for stress detection. To achieve this, they have induced stress in 40 volunteers through two cognitive tasks. The work, as a development of a BCI system for the automatic detection of cognitive states, is well-conceived. As the authors specify in their introduction, stress detection has been studied for many years; however, it is mostly conducted in controlled environments and with equipment that restricts or limits the free actions of experimental subjects. In this regard, I highly appreciate the effort to reduce the input variables to the BCI system to only EEG and ECG, as this minimizes the additional stress conditions induced by portable measurement systems.

Despite the efforts the authors have presented as innovative contributions of this study, I do not find significant contributions to the field. The choice of dry electrodes for EEG recording is consistent with the study's objective, as they are frequently used in portable devices. The choice of protocols to induce stress in the subjects may offer some novelty, such as the use of the video game “Paper, Please”, which has not been previously employed in similar tests. However, there are many recent studies that examine the cognitive impact of video game players using EEG signals.

The title of the paper should be changed to '...for stress monitoring.' While stress and motor and/or cognitive performance are related, the focus of this study should be directed towards stress.

In the Materials and Methods section, the authors describe the various traditional methods used to induce stress in laboratory experiments. This highlights the possibility of applying stress induction protocols through video game techniques. In particular, I find the approach of the proposed protocols interesting.

The description of the recording procedures and the features extracted from the ECG signals is adequate. However, for the EEG signals, it is understood that two bipolar signals were used, but it is not clear what features were extracted from these signals or what the machine learning model entails.

Sub-section 3.2 up to line 421 should be moved to the Materials and Methods section. I recommend relocating this sub-section to the Materials and Methods section, under the headings 'EEG Feature Extraction' and 'Computational Models for Stress Classification. Regarding the EEG features extracted, what was the size of the temporal segments used for their calculation?

In the discussion section, the authors present a series of comments related to the methods used for generating the stress state, as well as the limitations and future directions. This reduces the study to a mere implementation of machine learning for state classification. One way to enhance the results, discussions, and impact of this work could be to study the progressive dynamics of stress (if such a phenomenon exists). In their results, the authors showed that two blocks are sufficient to obtain the most optimal classifications. Could this result be used to assess stress classifications throughout the performance of cognitive tasks? That is, perform the classifications using two blocks at a time.

In its current state, the work does not represent a significant contribution to the neurosciences. In this context, the authors should leverage the classification results to explain the electrophysiological dynamics that could occur during tasks that induce stress conditions. For these reasons, I cannot recommend this work for publication in this journal.     

Minor

In line 310, it states, 'Below, Figure 4 shows an example of ECG and HRV recordings.' However, there is no ECG recording in this figure.

Line 334: Please correct the error in the reference.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Overall, the study is sufficiently detailed and well-structured General Remarks. The article is generally clear and well-structured, with each section logically building the narrative of the research. The description of the context and motivations is solid, In summary, the article provides a good overview of the project, with a well-described methodology and promising results.

Comments for author File: Comments.pdf

Author Response

Thank you for your response.

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript presents an EEG-based monitoring system to assess stress levels during cognitive tasks. Additionally, the authors introduce a signal decoding approach that interprets stress states from EEG recordings and evaluate its accuracy relative to ECG recording. The manuscript requires a major revision, as it is very difficult to read and lacks coherence. Additionally, the details within are not clearly presented and need better organization. Here are my comments:

1.     Title is too vague; include “stress-monitoring” related terms to improve clarity.

2.     The abstract lacks crucial details and leads to confusion: Does your BCI system incorporate both EEG and ECG data for decoding, or is the ECG data used separately for control purposes? Explicitly specify your CONTROL and the decoding method, rather than using a broad description such as "trained a BCI system."

3.     There is no detailed information about the EEG electrodes, including their specific placement coordinates and corresponding brain regions, how electrode positioning is maintained consistently across heads of different sizes, and the impedance levels for each patient before and after measurements.

4.     Even though citations are provided, it is still beneficial to include a detailed explanation of how entropy, fractal dimension, and Hjorth parameters are extracted (e.g. specific methods, calculations, and equations).

5.     Despite its mention on page 7 lines 252-253, Why is there no report on reaction times?

6.     Please have a thorough revision on the language, too many non-precisely word use. Please conduct a thorough revision of the language as there are too many places of imprecise wording.

Comments on the Quality of English Language

Please conduct a thorough revision of the language as there are too many places of imprecise wording.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

This new version of the manuscript shows a significant improvement. The materials and methods are now described in sufficient detail, and previous doubts regarding the procedures employed have been clarified and resolved. In light of the modifications made to the manuscript, I find the approach for detecting the progression of stress evoked by the proposed protocols both striking and interesting. Although I would like to see more details about these specific results, I understand that this is not the main focus of the study. Therefore, considering that the authors propose a methodology for stress detection, I believe this research could be a valuable contribution to those investigating real-time stress evaluation.

Brain-computer interface approaches, such as in this work, often do not fully explain the reasons behind the obtained classifications. However, they are a powerful tool for demonstrating that neural processes triggered by stress do exist and can be monitored at the cortical level. This study contributes to this understanding. In this context, I encourage the authors to continue their research, focusing on identifying the neural bases that allow machine learning technology to reveal differences in stress states. For example, why are the features used in this work relevant for the good performance of the models? Could it be that other features, such as frequency band energy or time-frequency characteristics, are not as effective, or are they?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have addressed my comments

Author Response

Thank you for your response.

Back to TopTop