Assessing the Impact of Sensor Orientation on Accelerometer-Derived Angles: A Systematic Analysis and Proposed Error Reduction

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study examined the impact of sensor orientation on accelerometer-derived angles. This study is intriguing and decently executed; however, some important improvements regarding the writing and explaining methods are needed. In particular, statistical analyses and discussion need to be addressed better. More specific comments are the following:

                                                                                                                                        

Introduction:

1.     Introduction is rather well written, with a good literature review and clear rationale. Some additional references are needed:

a.      Lines 47-56 can benefit from backing up these claims with adequate references.

b.     Same with lines 61-65

2.     Is there an option for some hypotheses?

3.     Finally, try to add the importance of the study at the end of the Introduction.

Methods:

1.     Line 97: Please add manufacturer information regarding Qualisys.

2.     Lines 117-119: Since the metronome was used, please elaborate on the velocity used.

3.     Same lines: Please provide some reference to the fact that this method with a metronome is adequate.

4.     Did the authors perform any statistical analyses? Please provide information and add adequate information regarding the statistical analyses.

Results:

1.     Line 168: Authors referring to ICC and excellent reliability. See my previous comment regarding the statistics.

2.     Line 186: Same as previous. Please provide adequate statistical analyses in the method section.

Discussion

1.     Lines 258-266 are not suitable for discussion but rather for introduction. Authors can refer to the previous studies in the discussion but not in this amount.

2.     Lines 272-279: Try to write concisely. These several sentences can be written as one sentence with two references.

3.     Lines 286-312 contain no references and primarily rely on speculating, assuming, expecting, could be…This is not a scientific manner of writing the discussion. Please rewrite accordingly.

4.     Add some limitations of this study.

Conclusions:

1.     Please add some practical applications and recommendations for further research.

 

Author Response

Reviewer 1

This study examined the impact of sensor orientation on accelerometer-derived angles. This study is intriguing and decently executed; however, some important improvements regarding the writing and explaining methods are needed. In particular, statistical analyses and discussion need to be addressed better. More specific comments are the following:

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted in the re-submitted files.                                                                                                                                      

Introduction:

1. Introduction is rather well written, with a good literature review and clear rationale. Some additional references are needed:

Thank you for your positive evaluation. We agree that adding references would strengthen the introduction and have added the following (see below under a. and b.)

1. Lines 47-56 can benefit from backing up these claims with adequate references.

Fisher, C.J., 2010. Using an accelerometer for inclination sensing. AN-1057, Application note, Analog Devices, pp.1-8.

Luinge, H.J. and Veltink, P.H., 2005. Measuring orientation of human body segments using miniature gyroscopes and accelerometers. Medical and Biological Engineering and computing, 43, pp.273-282.

1. Same with lines 61-65

Alsubaie, A.M., Sanderson, A., Cabral, H.V., Martinez-Valdes, E. and Falla, D., 2023. Spinal kinematic variability is increased in people with chronic low back pain during a repetitive lifting task. Journal of Electromyography and Kinesiology, 73, p.102832.

Sessa, S., Zecca, M., Lin, Z., Bartolomeo, L., Ishii, H. and Takanishi, A., 2013. A methodology for the performance evaluation of inertial measurement units. Journal of Intelligent & Robotic Systems, 71, pp.143-157.

2. Is there an option for some hypotheses?

We agree adding a hypothesis would strengthen this, therefore a hypothesis has now been included in the manuscript.

3. Finally, try to add the importance of the study at the end of the Introduction.

This has been added at the end of the introduction before the aim.

Methods:

1.     Line 97: Please add manufacturer information regarding Qualisys.

We have now added this information.

2. Lines 117-119: Since the metronome was used, please elaborate on the velocity used.

We agree with this addition and have added the necessary text and supporting reference.

 

3. Same lines: Please provide some reference to the fact that this method with a metronome is adequate.

See response above – additional references added.

4. Did the authors perform any statistical analyses? Please provide information and add adequate information regarding the statistical analyses.

Statistical analysis in the form of ICC for reliability and RMSE for peak and through range measures were included for both the accelerometer (corrected and un-corrected) and Qualisys. Furthermore, regression modelling was included. This can be found under data analysis and statistics section for Step 1 and data analysis and statistics for Step 2.

Results:

1. Line 168: Authors referring to ICC and excellent reliability. See my previous comment regarding the statistics.

See previous comment – clarification has been added in the methods section.

2. Line 186: Same as previous. Please provide adequate statistical analyses in the method section.

See previous comment – clarification has been added in the methods section.

Discussion

1. Lines 258-266 are not suitable for discussion but rather for introduction. Authors can refer to the previous studies in the discussion but not in this amount.

We agree with this suggestion and have moved this to the introduction and reanalysed it in the discussion.

2. Lines 272-279: Try to write concisely. These several sentences can be written as one sentence with two references.

Many thanks for identifying this. This paragraph has been reduced to two sentences.

3. Lines 286-312 contain no references and primarily rely on speculating, assuming, expecting, could be…This is not a scientific manner of writing the discussion. Please rewrite accordingly.

This section has been rewritten with the inclusion of more supportive evidence.

4. Add some limitations of this study.

A limitations section has now been added into the discussion.

Conclusions:

1. Please add some practical applications and recommendations for further research.

This has now been added into the conclusion.

 

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Editor,

Thank you for selecting me as a reviewer for the manuscript "Assessing the impact of sensor orientation on accelerometer-derived angles: A systematic analysis and proposed error reduction" submitted to Applied Sciences. The study presents an important investigation into how the initial orientation of accelerometers can introduce errors in derived joint angles. The authors systematically analyze this source of error across different offsets and movement velocities, and propose and test a mathematical correction method. The topic is novel and of significant interest given the increasing use of wearable sensors in biomechanics and movement analysis.

General comments:

 

Overall the study is well designed and presented. The aims, methods, results and interpretations are clear, and the paper reads logically with a robust methodology. The data analysis appears technically adept to achieve the stated aims. The main aspects of the current manuscript I would encourage the authors to address are:

• Include more background on previous related work in the introduction (range of agreement values reported, whether orientation was considered as a variable etc). This will help set the scene and rationale better.
• Explain if 20° and 50° tilt/twist offsets are typical of what might be seen in practice - i.e. are they reasonable clinically or ecologically relevant angles?
• Discuss movement amplitude as a potential variable - were different ROMs tested methodologically or might this influence findings?
• Comment on whether other joints/anatomical regions beyond the spine might also benefit from this data and proposed methods.
• Some additional discussion on limitations of the study should be added. For example, the use of a wood/metal jig rather than human joints.

 

Specific comments:

-              Expand on the specific aim outlined in Step 2 to correct error rather than just state error reduction.

-              References 21 and 22 demonstrate use of accelerometers to measure angles, but do you know if they corrected for orientation offsets? This would strengthen your point here.

-              Reference ranges for typical lumbar flexion/extension velocities would be useful for context when stating the range tested.

-              Provide some additional detail here on the fusion algorithms used in many IMU systems and their limitations.

-              In the data processing section, please provide more detail on the purpose written MATLAB algorithms used - what specific angle calculations or corrections were applied? Some brief pseudocode would help the reader understand better.

-              For the optoelectronic system, confirm the model, manufacturer, marker setup used to define segments and calculate angles.

-              Elaborate on the rationale for choosing a 1 Hz cut-off frequency for filtering. Lower cut-offs have also been reported so discuss pros/cons.

-              Specify if analysis was completed in custom written software or commercially available packages.

-              Figure 3: Panel labels A-C could be larger/clearer. Add units to the y-axis (degrees) Reduce x-axis clutter by removing every other value if percentiles not needed.

-              In Figure 3 and 4 legends, define abbreviations (RMSE) at first usage.

-              For table 1, suggestion to format R2 values to 2 or 3 decimal places for consistency.

-              In the discussion of results on page 6 (lines 185-189), comment whether the lower R2 values for frontal plane movements affect ability to use the predictive modeling equations accurately in these movements?

-              Comment on direction and clinical meaningfulness in differences between sagittal and frontal plan error increase per 10° change.

-              Consider labeling the different components in the image (e.g. IMUs, retroreflective markers) for clarity.

-              Consider rewording this sentence for clarity: "A third IMU, with a retroreflective marker attached, was used to time stamp and subsequently synchronize the accelerometer data with the OEM data."

-              The high ICC values demonstrate reliability is maintained across different initial orientations. This point should be highlighted in the results before presenting the effect on validity.

-              Quantify what constitutes "slow" vs "fast" angular velocities instead of just average and SD to aid interpretation.

-              In Figure 5, Increase font size of legend details. Reduce x-axis values for clarity as per Figure 3 suggestions.

-              In Figure 6, consider changing symbol/color for corrected data to better visually distinguish from uncorrected accelerometer data.

-              When discussing previous RMSE ranges reported in lines 261-263 from other papers, can you comment if those studies stated whether sensor orientation offsets were corrected for? This would give good context.

-              Suggest citing 1-2 more examples of relevant accelerometer validity/comparison to optoelectronics studies to show breadth in field beyond current 3 citations.

-              When discussing potential sources of error with reattachment of sensors (lines 269-272), can you estimate or cite expected sensor reattachment variations between sessions realistically? 5 degrees? 10 degrees?

Please address these comments to improve clarity and presentation of this manuscript. I believe this is a valuable study and addition to the literature if the authors can address these suggestions. 

 

Please do not hesitate to contact me with any questions.

Sincerely

Author Response

Thank you for selecting me as a reviewer for the manuscript "Assessing the impact of sensor orientation on accelerometer-derived angles: A systematic analysis and proposed error reduction" submitted to Applied Sciences. The study presents an important investigation into how the initial orientation of accelerometers can introduce errors in derived joint angles. The authors systematically analyze this source of error across different offsets and movement velocities, and propose and test a mathematical correction method. The topic is novel and of significant interest given the increasing use of wearable sensors in biomechanics and movement analysis.

Thank you very much for taking the time to review this manuscript. Please find the detailed

responses below and the corresponding revisions/corrections highlighted in

the re-submitted files.

 

 

General comments:

 

Overall the study is well designed and presented. The aims, methods, results and interpretations are clear, and the paper reads logically with a robust methodology. The data analysis appears technically adept to achieve the stated aims. The main aspects of the current manuscript I would encourage the authors to address are:

• Include more background on previous related work in the introduction (range of agreement values reported, whether orientation was considered as a variable etc). This will help set the scene and rationale better.

We agree this would strengthen the argument. Some additional text has now been included looking at the agreement offered in the literature.

• Explain if 20° and 50° tilt/twist offsets are typical of what might be seen in practice - i.e. are they reasonable clinically or ecologically relevant angles?

Some additional detail has been added about why 20 degrees was set, 50 degrees is justified in the text.

• Discuss movement amplitude as a potential variable - were different ROMs tested methodologically or might this influence findings?

Movement amplitudes used are in line 131 and the different peak values have now been added into the discussion.

• Comment on whether other joints/anatomical regions beyond the spine might also benefit from this data and proposed methods.

This has now been discussed from line 369 as part of the future contributions.

• Some additional discussion on limitations of the study should be added. For example, the use of a wood/metal jig rather than human joints.

This has now been added at the end of the discussion in a limitations section.

 

Specific comments:

-              Expand on the specific aim outlined in Step 2 to correct error rather than just state error reduction.

This has now been amended to clarify the aim of step 2 correcting error.

-              References 21 and 22 demonstrate use of accelerometers to measure angles, but do you know if they corrected for orientation offsets? This would strengthen your point here.

Many thanks for this comment. We agree that adding this would strengthen the argument. However, it is not clear form the text if corrections were completed or not, hence the speculative sentence following the reference.

 

-              Reference ranges for typical lumbar flexion/extension velocities would be useful for context when stating the range tested.

We agree adding this would strengthen the justification of the methods and have added references.

 

-              Provide some additional detail here on the fusion algorithms used in many IMU systems and their limitations.

Many thanks for this comment. This could benefit understanding, however limitations of IMUs has been included in the introduction briefly but a review of the specifics of difference fusion algorithms is beyond the scope of this manuscript.

 

-              In the data processing section, please provide more detail on the purpose written MATLAB algorithms used - what specific angle calculations or corrections were applied? Some brief pseudocode would help the reader understand better.

We agree that this could be clearer. The text has been modified and detail has been added.

 

-              For the optoelectronic system, confirm the model, manufacturer, marker setup used to define segments and calculate angles.

This has been clarified.

 

-              Elaborate on the rationale for choosing a 1 Hz cut-off frequency for filtering. Lower cut-offs have also been reported so discuss pros/cons.

This has been clarified with reference to the cited text.

 

-              Specify if analysis was completed in custom written software or commercially available packages.

This clarity has been added.

 

-              Figure 3: Panel labels A-C could be larger/clearer. Add units to the y-axis (degrees) Reduce x-axis clutter by removing every other value if percentiles not needed.

The graph axis has been tidied up. We believe the x axis marks are set to the offsets tested so are needed for clarity.  

 

-              In Figure 3 and 4 legends, define abbreviations (RMSE) at first usage.

These have been added into the figure captions now.

 

-              For table 1, suggestion to format R2 values to 2 or 3 decimal places for consistency.

Many thanks for identifying this inconsistency. All values in table 1 are now formatted to 4 decimal places for regression equations and 3 decimal places for R².

 

-              In the discussion of results on page 6 (lines 185-189), comment whether the lower R2 values for frontal plane movements affect ability to use the predictive modeling equations accurately in these movements?

This has been added into the discussion.

 

-              Comment on direction and clinical meaningfulness in differences between sagittal and frontal plan error increase per 10° change.

Thanks for bringing this to our attention, this discussion has now been added into the manuscript.

 

-              Consider labeling the different components in the image (e.g. IMUs, retroreflective markers) for clarity.

We agree with this suggestion and have created a new image.

 

-              Consider rewording this sentence for clarity: "A third IMU, with a retroreflective marker attached, was used to time stamp and subsequently synchronize the accelerometer data with the OEM data."

This has now been clarified and reworded.

 

 

-              The high ICC values demonstrate reliability is maintained across different initial orientations. This point should be highlighted in the results before presenting the effect on validity.

We agree with this point. This has been moved to earlier in the results section.

 

 

-              Quantify what constitutes "slow" vs "fast" angular velocities instead of just average and SD to aid interpretation.

This has now been added into the text, the figure caption and table notes.

 

-              In Figure 5, Increase font size of legend details. Reduce x-axis values for clarity as per Figure 3 suggestions.

This has now been completed for figure 5.

 

-              In Figure 6, consider changing symbol/color for corrected data to better visually distinguish from uncorrected accelerometer data.

The colours have now been changed to allow greater clarity between the data types.

 

-              When discussing previous RMSE ranges reported in lines 261-263 from other papers, can you comment if those studies stated whether sensor orientation offsets were corrected for? This would give good context.

This has been added where appropriate.

 

 

-              Suggest citing 1-2 more examples of relevant accelerometer validity/comparison to optoelectronics studies to show breadth in field beyond current 3 citations.

More relevant studies have been added to the validity between sensor and Optoelectronic comparison.

 

-              When discussing potential sources of error with reattachment of sensors (lines 269-272), can you estimate or cite expected sensor reattachment variations between sessions realistically? 5 degrees? 10 degrees?

Many thanks for this comment. It is something that is not addressed in the literature. We plan some future work exploring this. It would be highly likely that reattachment would result in 10+ degrees tilt. 

Please address these comments to improve clarity and presentation of this manuscript. I believe this is a valuable study and addition to the literature if the authors can address these suggestions. 

 

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for addressing all the comments and raised issues.

Best regards