Biomarker-Informed Machine Learning Model of Cognitive Fatigue from a Heart Rate Response Perspective

Round 1

Reviewer 1 Report

The authors present a review and position paper on machine learning informed approaches to developing biomarkers for assessing the psychological state of cognitive fatigue. The paper is well written, the key messages are clearly conveyed, and the authors make an important contribution to the field of cognitive fatigue research. Minor revisions will likely improve the manuscript so that it is acceptable for publication:

• The authors provide a helpful critical evaluation of the existing self-report and objective measures of cognitive fatigue, but it would also be helpful to mention the relationship between self-reported and performance based measures of cognitive fatigue - are these typically related to one another? Also, is there any evidence that multi-tasking cognitive batteries elicit more reliable assessments of cognitive fatigue that typical single task measures?
• It would be worth mentioning why the frequency domain as opposed to time or non-linear domain components of HRV have been the focus in the cognitive fatigue literature
• Line 164: "these findings indicate that the sympathetic..." should also include mention of the phrase 'nervous system' or activity to improve clarity
• Line 182:  the authors should define "sympathovagal"
• It would be helpful to mention how HRV has been measured in the previous studies and some of the limitations of these approaches
• Table 3 contains errors such as 'postive' instead of 'positive

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Manuscript ID: sensors-1203200

Manuscript title: Biomarker-Informed Machine Learning Model of Cognitive Fatigue from a Heart Rate Response Perspective

COMMENTS:

General:

The work provides a review on recent advances in cognitive fatigue assessment with a particular focus on: heart rate variability indices proposed as cognitive fatigue biomarkers; machine learning approaches for building predictive models of cognitive fatigue.

The overall manuscript is well-written and fits the scope of the journal. Findings are interesting and well discussed, with particular regard to the section “Issues and Implications”.

However, I believe that some revisions are needed before considering it for publication, as outlined in the following lines.

Lines 53-59:

Here the authors outline the structure of the paper. In my opinion, authors could better describe and clarify the methodological approach and justification of the study, which is only very briefly reported in the ending lines of the Introduction section. Even though this is not a systematic review, I believe it would be worth highlighting how the review has been carried out, which timespan has been considered, how many studies have been examined, etc. This would give an idea of the workflow adopted to conduct the research and it would provide a measure of the size of the proposed study as well as its methodological skeleton.

Lines 63-64:

I would suggest proving a general description of the traditional psychological approaches to cognitive fatigue assessment before commenting on their pros and cons. In particular, I believe that it would beneficial to the readers to have a brief introduction to the traditional methodologies (what are their basic principles/assumptions, how are they typically carried out, etc…) before examining their possible advantages and limitations.

Lines 92-93:

Please provide a definition of the terms “subjectivity”, “disruptiveness”, “timeliness”, and “generalisability”, giving a reason why they have been selected. Although they are briefly explained at the end of Table 1, I would suggest adding a description in the main text and also explaining why the authors chose such parameters for the evaluation of self-assessment and task performance methods for cognitive fatigue evaluation.

Lines 141-142:

Before introducing the Table 2, it would be worth describing at least the main scope and main differences between time-domain, frequency-domain, and non-linear HRV indices. This would guide the reader towards a better understanding of the Table 2.

Lines 213-215:

It should be mentioned that the accuracy of regression models is mainly determined by the R-square value as well as by the MAE, MSE, and RMSE.

Lines 318-319:

This sentence is supported by valid but old references. It could be strengthened by adding more recent references to further validate the statement about the risk of misinterpretation of machine learning models.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf