Can Machine Learning Predict Running Kinematics Based on Upper Trunk GPS-Based IMU Acceleration? A Novel Method of Conducting Biomechanical Analysis in the Field Using Artificial Neural Networks
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsGeneral Comments
The aim of this study was to investigate whether running kinematics can be accurately estimated through an artificial neural network model containing GPS-based accelerometer variables, and anthropometric data.
I thank the authors for the time taken to correct and resolve the various corrections. Most of the problems have been solved and the English has been improved. There are some minor considerations that need to be addressed. These considerations are included in the specific comments.
I kindly ask the authors to read this report carefully and to respond accurately to the suggestions made if they consider them appropriate. I thank them for their time and investment in this paper.
Specific Comments
Abstract
Shorter sentences in the abstract. Difficult to read
Eliminate repeated keywords in the title. They keep appearing
Introduction
The introduction still contains excessively long sentences. In English, the writing should be more concrete, with shorter sentences that make it easier to read.
The introduction needs to be better structured, to relate the ideas in the paragraphs and to be more concrete in their application. It is difficult to read.
Material and methods
To better define the criteria for inclusion and exclusion of participants. Simply stating that they are experienced is not enough.
Add validation studies of the different materials used.
Add references to support the reasons for the measures used.
For the output variables it would help a lot to add a table instead of the distribution that has been made.
Discussion
This section needs to be rewritten. Confusing and shallow. No linkage of ideas to previous studies and rarely any causal relationship possible. Needs to be clearer and related to studies in the research field.
Conclusions
More concrete. Do not repeat the object of study.
Comments on the Quality of English LanguageRevision of the entire text. Difficult to read, unclear.
Author Response
General Comments
The aim of this study was to investigate whether running kinematics can be accurately estimated through an artificial neural network model containing GPS-based accelerometer variables, and anthropometric data.
I thank the authors for the time taken to correct and resolve the various corrections. Most of the problems have been solved and the English has been improved. There are some minor considerations that need to be addressed. These considerations are included in the specific comments.
I kindly ask the authors to read this report carefully and to respond accurately to the suggestions made if they consider them appropriate. I thank them for their time and investment in this paper.
Reply:
Thank you for taking the time to review our manuscript. We greatly appreciate the feedback provided. We have made several amendments to the manuscript. Particularly, with the style of English used within the manuscript. We have also included reference to our previous study which has now been published. With this, the rationale for the methodology used should be much clearer. Please see below for responses to the specific comments. All amendments have been highlighted in red text throughout the manuscript.
Specific Comments
Abstract
Shorter sentences in the abstract. Difficult to read.
Reply:
Thank you pointing this out. We have now shortened some of longer sentences in the abstract to improve clarity and readability.
Eliminate repeated keywords in the title. They keep appearing.
Reply:
These are now revised accordingly. Please see the revised title and Keywords.
Introduction
The introduction still contains excessively long sentences. In English, the writing should be more concrete, with shorter sentences that make it easier to read.
Reply:
Thank you for this feedback. We have now restructured the introduction and made the sentences shorter and more concise throughout the Introduction to improve clarity and readability.
The introduction needs to be better structured, to relate the ideas in the paragraphs and to be more concrete in their application. It is difficult to read.
Reply:
Thank you for this feedback. We have now restructured the introduction and added further text to help explain the ideas in each paragraph together with the connecting sentences to ensure flow throughout the section. The structure is as follows:
Paragraph 1: Introduction into IMUs.
Paragraph 2: How previous studies have employed IMUs to measure kinematics before the integration of ANNs. Including a description of how sensor fusion methods utilise IMU data.
Paragraph 3: Further information into the accuracy of sensor fusion methods and the limitations of employing these methods in the field.
Paragraph 4: Introduction into the benefits of newer methods that use ANNs and how these overcome previous issues.
Paragraph 5: Further information into the different ANNs that have been employed in estimating kinematics. Including their accuracies in walking and running analysis.
Paragraph 6: Explanation of the use of IMUs within sports. Explanation into the rationale of the current study and how the GPS-based IMU can potentially measure kinematics.
Paragraph 7: Explanation of how utilising other variables available to sports science practitioners within ANN could strengthen their ability to predict running kinematics.
Paragraph 8: Rationale for conducting kinematic analysis with GPS-based IMU with athletes. Also including the aim and intentions of the present study.
Material and methods
To better define the criteria for inclusion and exclusion of participants. Simply stating that they are experienced is not enough.
Reply:
Thank you for pointing this out. Further information is now added to the manuscript. Please see the additional information provided in lines 133-135.
Add validation studies of the different materials used.
Reply:
Thank you for mentioning this. We have now provided further information to the sampling frequencies used (lines 145 & 152). These would add to the references for the motion capture system and marker set (line 150) that we had already provided.
Add references to support the reasons for the measures used.
Reply:
Thank you we have now included rationale and reference to our previous study which help explain this. This is in the Introduction (lines 98 – 100 & 105-106) and also in the Methodology (lines 124 – 125, 161 – 163 & 176 – 179) sections.
For the output variables it would help a lot to add a table instead of the distribution that has been made.
Reply:
Thank you, another reviewer also mentioned this. We have now included all output variables in the form of a table (Tables 1 & 2, page 5)
Discussion
This section needs to be rewritten. Confusing and shallow. No linkage of ideas to previous studies and rarely any causal relationship possible. Needs to be clearer and related to studies in the research field.
Reply:
Thank you for this feedback. We have now restructured paragraphs within this section to ensure further in-depth discussion of the results of this study in light of the literature and expanded on the practical implications. Also in each paragraph, we now give an explanation or concluding remarks on the findings in the last few sentences. We also have now made comparisons to previous studies (lines 292 – 297, 316 – 324). Also, there are now two paragraphs on how the findings could be used within practice (Lines 329 – 353) which we added. Lastly, we have now included future directions in the final paragraph (lines 367-373).
Conclusions
More concrete. Do not repeat the object of study.
Reply:
Thank you we have now restructured the conclusion and added in new text to create a more robust concluding remarks.
Reviewer 2 Report
Comments and Suggestions for Authorsapplsci-2861102-peer-review-v1
This manuscript presents a study on running kinematic estimation using an ANN model containing GPS-based accelerometer variables and anthropometric data.
Title
The use of “AI” is ambiguous. Please revise the title to reflect on the applied method more accurately.
Methods
The authors emphasized the use of the ANN model and anthropometric data in running kinematic estimation. I did not find any results regarding participant anthropometric data or how the authors normalized or applied the anthropometric data in the data transformation/data analysis.
Line 149. Please provide participants’ demographic data.
Section 2.3.1. Please provide a schematic of the angles and joint segments.
Section 2.3.2. The results are summarized in table and hard to interpret their similarities and differences. It would be great if the authors could provide visual aids in running cycle/phase data between the IMU and ANN.
Discussion
The authors mainly focused on reporting the RMSE and rRMSE results separately from Figures 1 and 2, respectively. The discussion would be significantly strengthened by a more in-depth interpretation of these results, particularly considering the notable individual differences in the measurements highlighted by the figures. Such an analysis would provide a more comprehensive understanding of the implications and findings of the study.
Minor comment.
Line 143 "(Chapter 2)" has been mentioned without any subsequent explanation. This reference is unclear and can potentially disrupt the flow of the manuscript. If this manuscript is based on the author's own dissertation, it is important that it be correctly formatted.
Author Response
This manuscript presents a study on running kinematic estimation using an ANN model containing GPS-based accelerometer variables and anthropometric data.
Title
The use of “AI” is ambiguous. Please revise the title to reflect on the applied method more accurately.
Reply:
Thank you for this feedback. The title was also mentioned by another reviewer's comments and we have now revised the title to more specifically reflect our work.
Methods
The authors emphasized the use of the ANN model and anthropometric data in running kinematic estimation. I did not find any results regarding participant anthropometric data or how the authors normalized or applied the anthropometric data in the data transformation/data analysis.
Reply:
Thank you for this comment. The anthropometric data was used as an independent input variable. So therefore, we did not use anthropometric data to normalise any of the other input variables. We have now provided further information to describe the settings used within the ANN (section 2.4) for further clarification. In addition, the reason why we did not provide specific anthropometric data in the result is due to the outputs of the ANNs. The ANN combined the values of all input variables to estimate the output variable. Therefore, the only information that can be provided is the model summary (i.e. how accurate it was) and the relative importance of each input variables within the model.
Line 149. Please provide participants’ demographic data.
Reply:
Thank you for this comment. We have now included the participants height and mass along with their sex and age (line 132). We have also now included further information into the inclusion criteria (line 133).
Section 2.3.1. Please provide a schematic of the angles and joint segments.
Reply:
Thank you for this suggestion. We also now added a Schematic (Figure 1) to present the results, and think that can help as a visual aid with understanding the results as per next comment. This has now facilitated a visual understanding of the variables included in the model. Another reviewer also commented on the structure of the output variables in the methods section and ask for a table to be added. Considering there are 49 joint/segment kinematics a table was used. However for the most important variables the reader was referred to consult Figure 1 (page 10). Please see tables 1 & 2 (page 5).
Section 2.3.2. The results are summarized in table and hard to interpret their similarities and differences. It would be great if the authors could provide visual aids in running cycle/phase data between the IMU and ANN.
Reply:
Thank you for pointing this out. We have now included a figure with the most accurate results from the ANNs in relation to the gait cycle (Figure 1, Page 10).
Discussion
The authors mainly focused on reporting the RMSE and rRMSE results separately from Figures 1 and 2, respectively. The discussion would be significantly strengthened by a more in-depth interpretation of these results, particularly considering the notable individual differences in the measurements highlighted by the figures. Such an analysis would provide a more comprehensive understanding of the implications and findings of the study.
Reply:
Thank you for this suggestion. We agree that including all the results of the ANNs would provide greater analysis. We have therefore done this in the results section (Figure 1) concerning the most accurate output variables. Also we now have included reference to the sensitivity analysis in the discussion to further emphasise the findings relating to anthropometric data (lines 240 – 246 & 324 – 326)
Minor comment.
Line 143 "(Chapter 2)" has been mentioned without any subsequent explanation. This reference is unclear and can potentially disrupt the flow of the manuscript. If this manuscript is based on the author's own dissertation, it is important that it be correctly formatted.
Reply:
Thank you pointing this out. This was a mistake and we have now rectified it. In addition, our previous study has now been published which we have now made references to throughout the present study. Hopefully this will provide more context for the reader regarding the rationale behind the study and the variables chosen.
Thank you for taking the time to review this manuscript and we greatly appreciate your feedback.
Reviewer 3 Report
Comments and Suggestions for Authors1. In the methodology section, it is recommended to provide more detailed information on the models and technical specifications of GPS devices, accelerometers, and motion capture systems.
2. When describing the input variables of a neural network model, it is recommended to provide more detailed explanations, especially regarding why these variables were selected and how they reflect running kinematics.
3. Method section: It is recommended to provide more detailed information on the steps and parameter settings for calculating MLP ANNs using IBM SPSS Modeler software.
4. When using professional terminology, it is recommended to provide clear explanations to ensure that the terms in the text are understandable to the general reader.
5. It is suggested that the author further prospects future research directions or application prospects in practical sports science and medical practice, in order to make the conclusions more profound and practical
Comments on the Quality of English LanguageMinor editing of English language required.
Author Response
- In the methodology section, it is recommended to provide more detailed information on the models and technical specifications of GPS devices, accelerometers, and motion capture systems.
Reply:
Thank you for taking the time to review this manuscript and we greatly appreciate your feedback. We have now provided further information into the accelerometer settings and the set up of the device (see lines 143 – 152).
- When describing the input variables of a neural network model, it is recommended to provide more detailed explanations, especially regarding why these variables were selected and how they reflect running kinematics.
Reply:
Thank you for this comment. The rationale for the inclusion of the variables was mainly based on our previous study that has now been published. We have bow made references to that throughout the current manuscript. This is in the introduction (lines 98 – 100 & 105-106) and also in the methodology (lines 124 – 125, 161 – 163 & 176 – 179) sections.
- Method section: It is recommended to provide more detailed information on the steps and parameter settings for calculating MLP ANNs using IBM SPSS Modeler software.
Reply:
Thank you pointing this out. We have now provided further explanations into the settings used within the ANN and to why these were chosen (lines 194 – 202).
- When using professional terminology, it is recommended to provide clear explanations to ensure that the terms in the text are understandable to the general reader.
Reply:
Thank you again, we have provided a more detail description of the ANNs and hopefully this helps to explain the different components (lines 194 – 202).
- It is suggested that the author further prospects future research directions or application prospects in practical sports science and medical practice, in order to make the conclusions more profound and practical.
Reply:
Thank you pointing this out. We have now included an extra paragraph at the end of the discussion to suggest further directions based off our findings where the practical implications are further highlighted (lines 367 – 373).
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThank you very much for your considerations, I think it has improved the level of the paper considerably. I have nothing more to say.
Author Response
Thank you again for your feedback in this process.
Reviewer 2 Report
Comments and Suggestions for AuthorsThank you for addressing my comments.
Please carefully check for typography and formatting errors.
Author Response
Thankyou again for your feedback in this process. I have proof read the document again now and made more grammatical amendments.
