Two-Step Validation of a New Wireless Inertial Sensor System: Application in the Squat Motion

Round 1

Reviewer 1 Report

The manuscript titled ‘Two-steps validation of a new wireless inertial sensor system : application in the squat motion’  proposes a newly designed IMU system for clinical assessment and rehabilitation follow-up of patients with low back pain (LBP) validated through two main methods, one based on purely sensor validation, the other based on segmental functional validation.   

The work is interesting, but some aspects should be improved by the authors for considering the manuscript for a possible publication:

• Line 102: the ‘number of marks (13) should not include the markers placed over IMUs; please specify it better in the text;
• In Figure 2, for clarity, perhaps it should be better to nominate the IMUs with different symbols (i.e. IMU 1, IMU 2, etc).
• Line 128: you write that the data from IMU and VICON systems were not synchronous and that a cross correlation algorithm based on linear acceleration was developed to resynchronize3d it. Could you please specify better this methodology?
• Line 185 when you write that small vibrations of the hexapod platform are captured by accelerometers measurements but not by VICON system, what does it mean? Are the vibrations real? Do they produce a so high LOA (-11.597)?
• Paragraphs 4.3 and 4.4 have formatting errors and incomprehensible question marks and they are illegible at present.
• Conclusions are too short and they need to be extended.

Author Response

All the authors thank kindly the reviewers for their constructive feedbacks.

• Line 102: the ‘number of marks (13) should not include the markers placed over IMUs; please specify it better in the text;

Two modifications were performed

1. We added the sentence ‘Clusters of 3 markers were placed on each IMUs to compute their orientation via the optoelectronic system.’
2. We changed the marker numeration in figure 2, placing the IMU marker number at the end

• In Figure 2, for clarity, perhaps it should be better to nominate the IMUs with different symbols (i.e. IMU 1, IMU 2, etc).

Figure 2 has been updated with number on IMU

• Line 128: you write that the data from IMU and VICON systems were not synchronous and that a cross correlation algorithm based on linear acceleration was developed to resynchronize3d it. Could you please specify better this methodology?

We added the following description in appendix section.

            As explained, the VICON and IMU, being two independent measurement systems, are not synchronized. One solution for this problem could have been to start and stop both systems at the same time but this method was not chosen for it is not precise enough and does not prevent the risk of a constant delay between measurements. Hence the cross-correlation procedure was chosen to reduce the risk of delayed data.

The algorithm is composed of the following steps:

1. Computing the average point of the back IMU cluster in VICON
2. Double derivation of the trajectory to obtain linear acceleration (accel_vic)
3. Cross correlation (xcorr MATLAB build-in function) of the accel_vic with accel_imu, i.e. the computed linear acceleration of the back IMU
4. Identification of the cross-correlation maximum value r_max
5. Synchronization of the IMU and VICON signals at r_max time
6. Cropping of the IMU and VICON data to obtain similar length for both signals.

All computations were programmed in MATLAB R2020b software.

• Line 185 when you write that small vibrations of the hexapod platform are captured by accelerometers measurements but not by VICON system, what does it mean? Are the vibrations real? Do they produce a so high LOA (-11.597)?

This line has been moved to section 4.1 because it is by nature a limitation discussion

As can be seen on Figure 1, the hexapod platform is hollow at the center, which means that fixation of the wooden casing can only be performed on the lateral extremities. Hence, bending of the casing is permitted and a potential vibration generator if the casing hits the platform.

The typical dynamic observed on Figure 3-b can be explained by elastic deformation of the wooden casing during the acquisition.

Hence we present 2 justifications which could impact the accuracy of the system

1. The works of Lebel et al 2013 showed that increasing the angular velocity reduced the accuracy of the IMU. Hence increasing the hexapod rotation during the acquisition may have affected the accuracy of our sensors
2. The hexapod platform and the wooden casing may have introduced some artefact accelerations in our measurements.

• Paragraphs 4.3 and 4.4 have formatting errors and incomprehensible question marks and they are illegible at present.

We are deeply sorry to read this review. Both WORD and PDF files appears clearly in our computer. Moreover, other reviewers do not raise a similar issue.

• Conclusions are too short and they need to be extended.

Extensions have been made

Reviewer 2 Report

see joint file

Comments for author File: Comments.pdf

Author Response

All the authors thank kindly the reviewers for their constructive feedbacks.

1. In the abstract, the authors mention that the main objective of their study is to validate a newly designed IMU system; However, they didn’t give any information about that system in the following sections of the paper.

The reviewer is right, as explained in section 2.2, the details of the IMU system have already be presented in previous work. The authors chose to avoid a repetition by making a direct reference to the article

Hubaut, R.; Guichard, R.; Greenfield, J.; Blandeau, M. Validation of an Embedded Motion-Capture and EMG Setup for the Analysis of Musculoskeletal Disorder Risks during Manhole Cover Handling. Sensors 2022, 22, 436, doi:10.3390/s22020436.

For more details, here is the description:

The motion-analysis system designed within the scope of consortium INTERREG: NOMADe, is composed of a Data Capturing Unit (DCU) (Dramco KUL, Ghent, Belgium) receiving the measured data of four wireless IMUs (Dramco KUL, Ghent, Belgium) (see Figure 1). The DCU acts as a controller (IMU settings and acquisition START/STOP), a preprocessor (IMU calibration and fusion of the raw data via a Madgwick filter to obtain the orientation quaternion) and storage of the acquired data. The four IMUs sensors are based on a very well-known Micro Electro Mechanical Systems (MEMS) IMU: the MPU6050, which combines a 3-axis gyroscope with a sensitivity range from 250 to 2000°/s and a 3-axis accelerometer with a sensitivity range from ±2 to ±16 g. The sample frequency can be set to 10, 20, 25, 50, or 100 Hz. The quaternions, accelerometer, and gyroscope data communicated through Bluetooth are stored in a text format file in a micro-SD card on the DCU.

2. In page 4, the authors didn’t explain why the sampling frequency of the IMU was set to 50Hz.

The reviewer is right, as explained in section 2.4.1, the IMU system is still new and currently in continuous improvement process. One major update involves Bluetooth connection with the IMU which can suffer from packet loss, especially when using high frequency. For our study, we found that 50Hz was a good compromise between packet loss and frequency quality.

Moreover, the frequency is in coherence with the motion of interest. We measured between 7 to 13 squats per acquisition of 10secondes duration Hence 50Hz is sufficient to respect Nyquist frequency.

3. Again, why the sampling frequency of the VICON was set to 100Hz? is it a characteristic of the system or it is a required one?

The 100Hz frequency is the default set-up for VICON system. We added this information in the manuscript.

4. In section 2.4.1, the authors used a 4th order Butterworth low-pass filter and they mention two cut-off frequencies! is it a band-pass filter? or they used two low-pass filters? and they should explain why did they choose these cut-off frequency?

The two cut-off frequency of 6Hz and 20Hz are respectively for VICON signals and IMU signals. Both frequency were chosen based on scientific literature and the references are specified.

We changed the sentence to improve clarity.

5. some abbreviations are not given.

Those abbreviations have been corrected

6. Based on the four equations, and using the standard definition of the Conjugation of quaternion and its property, can we conclude that:

The reviewer is right, this is another way these equations.

7. In section 2.5, do the authors mean Lin’s concordance correlation coefficient (CCC)?

The reviewer is absolutely right, this is a translation mistake and has been changed in the manuscript.

8. The results shown in the two tables should be better explained.

The methodology and all parameters computed in table 2 and 3 have been detailed in section 2.5

9. In Table 2, for upper LOA, why the observed value is almost the double of that of the reference?

We are sorry but we do not understand this remark, could the reviewer please detail what is the observed value and the reference?

10. In figure 3 - a, many points are above or below the green line!! the authors mention that for some security limits the VICON system can not record the results, can they conduct other experiments to highlight this effect or to avoid it?

The reviewer is right, some measured points are outside of the LOA defined in the Bland-Altman methodology. We explain this phenomenon by two reason:

1. the internal system vibration which were discovered after the acquisition. Unfortunately, we do no longer have access to the hexapod and cannot perform other experimentation on it yet.
2. the statistical definition of LOA being the bias value (mean of difference between systems) plus-or-minus 1.96 times the standard deviation of difference between systems. Hence the longer the acquisition, the more chance there are to observe outsider values on this plot.

11. In figure 3 - b, it seems that the linear regression is not the best approximation, by looking at the point we may expected a polynomial with 4th degree will be more suitable to fit these points?

The reviewer is right, on this specific example of Bland-Altman plot, it seems the acceleration does not linearly fit the 45° line. This typical dynamic can be explained by elastic deformation of the wooden casing during the acquisition.

Following recommendation from Bland & Altman 1986, it makes more sense to keep comparing the systems through a linear regression for they both measure the same thing: linear acceleration.

Statistical methods for assessing agreement between two methods of clinical measurement J. Martin Bland, Douglas G. Altman, 1986, The Lancet.

12. Similar comments for Figure 4.

The same answers than is reviews 10 and 11 apply here.

We also add the following argument the problem of error increasing with the angular velocity can be observed on Figure 4 and has been discussed in Taylor et al 2017 and its impact on orientation estimations will be very low considering the angular velocities observed in real clinical tasks.

Taylor, L.; Miller, E.; Kaufman, K.R. Static and Dynamic Validation of Inertial Measurement Units. Gait Posture 2017, 57, 80–84, doi:10.1016/j.gaitpost.2017.05.026.

13. In Table 3, why there are big difference among the values given by the two systems?

As discussed in section 4.4, the differences are not so important compared to previous scientific literature. Moreover, those differences must be compared relatively to the range of flexion/extension during a squat which can be around 90°.

14. In section 4.1, the IMU accuracy is more affected by the velocity or the acceleration?

We advance two limitations arguments in 4.1.

1. The works of Lebel et al 2013 showed that increasing the angular velocity reduced the accuracy of the IMU. Hence increasing the hexapod rotation during the acquisition may have affected the accuracy of our sensors
2. The hexapod platform and the wooden casing may have introduced some artefact accelerations in our measurements.

Lebel, K.; Boissy, P.; Hamel, M.; Duval, C. Inertial Measures of Motion for Clinical Biomechanics: Comparative Assessment of Accuracy under Controlled Conditions - Effect of Velocity. PLOS ONE 2013, 8, e79945, doi:10.1371/journal.pone.0079945.

15. In line 221, why will the wooden casing generate vibrations?

As can be seen on Figure 1, the hexapod platform is hollow at the center, which means that fixation of the wooden casing can only be performed on the lateral extremities. Hence, bending of the casing is permitted and a potential vibration generator if the casing hits the platform..

16. How can the authors verify the statement given in lines 226 and 227?

Soft tissue artefact (STA) is still an open problem in biomechanical acquisition. The best way to assess the impact of STA is to perform motion capture with either a medical imaging techniques like double x-ray fluoroscopy or markers directly implanted in the bones via intracortical screws.

Without relying on such complex materials and method, we published in previous work a preliminary study on the variability of thigh STA and wobbling depending on the position of the sensors.

Guichard, R.; Blandeau, M.; Leteneur, S. Localization of IMU Sensors Affects the Estimation of Soft Tissue Wobbling: 412 A Preliminary Study. Comput. Methods Biomech. Biomed. Engin. 2021, 24, 3

17. Can the authors give more information or hints to validate the conclusion given in line 233.

This sentence being here is a regrettable mistake and for it belonged to an old version of this article. It has been removed.

18. what is the purpose of section 4.2, as it contains generic and known information?

The purpose of this section is to clarify on our technical terminology for reliability and agreement as previous works showed discrepancies in the use and interpretation or these core concepts.

Kottner, J.; Streiner, D.L. The Difference between Reliability and Agreement. J. Clin. Epidemiol. 2011, 64, 701–702, doi:10.1016/j.jclinepi.2010.12.001.

19. In section 4.3, there are many typos.

The manuscript has been read thoroughly by all authors and by a native English speaker colleague (J. Greenfield).

20. What are the major difference between figures 3 and 4? The caption of the two figures should be rewritten.

Bland-Altman plot can only be created for one parameter

The first Bland-Altman plot is for linear acceleration, the other for angular velocity

Caption have been rewritten to improve clarity.

21. The authors mention that their 47th cited reference conducted similar experiments? so what is the main difference between the proposed study and his study?

The work from Lepetit 2019 focus on the evaluation of a specific clinical test (Timed up and Go) through the use of one IMU placed on the chest.

One main difference in his validation method is he did not go through the 2 steps we currently present in our article.

22. The conclusion should be further developed.

Extensions have been made

23. The authors should avoid when possible to cite some non-English references that many readers can’t read them.

We chose to cite the French works of Lepetit 2019 on the 47^th cited reference and not his other English speaking articles because there were not focused on the acceleration signal but on the kinetic energy. Yet we also referenced the English work on the 30^th cited reference.

Lepetit K, Ben Mansour K, Boudaoud S, Kinugawa-Bourron K, Marin F. Evaluation of the kinetic energy of the torso by magneto-inertial measurement unit during the sit-to-stand movement. J Biomech. 2018 Jan 23;67:172-176. doi: 10.1016/j.jbiomech.2017.11.028.

24. Some variables and operators are not defined or well defined.

Corrections have been made, especially in section 2.5

25. The manuscript needs extensive revision for language.

The manuscript has been read thoroughly by all authors and by a native English speaker colleague (J. Greenfield).

26. There are few typos in the manuscript.

The manuscript has been read thoroughly by all authors and by a native English speaker colleague (J. Greenfield).

27. The authors of this manuscript should emphasize their contributions.

This is specified in the Author Contributions: section

28. Some parts of the manuscript are overstressed

Could the reviewer please precise which part of the manuscript?

Reviewer 3 Report

This study is to explore two-steps validation of a new wireless inertial sensor system with 2 application in the squat motion. The study motivation is clear and study purpose are proper. Overall study processes are valid and sound. 

Followings are some comments to improve the quality of the paper. 

• In line 74, enrich the study purpose .
• In line 83, if possible, provide separate statistic for female and males. three females data among 15 subjects may cause a big bias. 
• In line 185, if ab abbreviation shows the first time in the paper, provide a full name; Limits of Agreement (LOA). 
• In line 189, confidence intervals may be provided in Table 2, if available. 
• If applicable, provide t-test statistic between two measurement methods and ANOVA test statistic among different motions.

Author Response

All the authors thank kindly the reviewers for their constructive feedbacks.

• In line 74, enrich the study purpose .

Changes have been performed in the introduction to improve clarity.

• In line 83, if possible, provide separate statistic for female and males. three females data among 15 subjects may cause a big bias. 

Changes have been performed

• In line 185, if ab abbreviation shows the first time in the paper, provide a full name; Limits of Agreement (LOA). 

The reviewer is right; this has been corrected

• In line 189, confidence intervals may be provided in Table 2, if available. 

We discussed the problem of error increasing with the angular velocity on Figure 4. Unfortunately, this phenomenon induces a non-normal distribution of error making it impossible to compute confidence intervals.

This error has been discussed in Taylor et al 2017 and its impact on orientation estimations will be very low considering the angular velocities observed in real clinical tasks.

• If applicable, provide t-test statistic between two measurement methods and ANOVA test statistic among different motions.

Unfortunately, it is not applicable in our situation, we do not have enough varying parameters to use ANOVA statistical test. Future works could assess through this methodology the effect of some squatting strategies or subjects anthropometric size or gender on the precision.

Round 2

Reviewer 1 Report

Ok, you answered to the coments improving the manuscript quality. There is one problem of visualitation (many typos) as you can see from the attached detail. Please check it.

Comments for author File: Comments.docx

Author Response

All authors kindly thank the reviewer for the feedback.

The typos have been corrected via the MDPI English editing service.

Reviewer 2 Report

The second version of the manuscript is much better than the previous one.

This version of the manuscript still needs extensive revision for language. For examples: line 21, “has” should be replaced by “have”, same problem at the 24^th line, “suffer” in line 28^th etc.

There are few typos in the manuscript. Please check the pdf file.

Author Response

All authors kindly thank the reviewer for the feedback.

The typos have been corrected via the MDPI English editing service.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This work presented an inertial measurement unit for squat motion. The structure of the device was relatively simple and the method adopted here was facile. Some experiments showed that this method was beneficial. Please consider the following suggestions.

(1) The authors should provide diagrams of how the device works, which will give the potential readers a better idea of how the device works.

(2) The authors should clearly point out the innovation of this work. It is also necessary to compare the performance of other devices.

Author Response

All the authors thank kindly the reviewer for the constructive feedbacks.

(1) The authors should provide diagrams of how the device works, which will give the potential readers a better idea of how the device works.

The figure 1 has been updated

(2) The authors should clearly point out the innovation of this work. It is also necessary to compare the performance of other devices.

Pointed out at line 73 “To the best of our knowledge, we found no study combining those three approaches.”

In the discussion section, performances are indeed compared against various other devices, for instance, Lepetit et al. 2019 used and APDM Opal (Portland, USA) and Lebel et al. 2013 used APDM Opal, Xsens MTx Inertial Lab, OSv3.

Reviewer 2 Report

This manuscript presents a validation methodology for a new wireless inertial sensor system. The proposed validation is interesting, but:

Equation 3 and 4 seems not correct (Hamilton product, or simple product?)

The regression and the squat detection algorithm are not described in detail in the manuscript.

The magnetometer is an important sensor that should also be used in 3D body orientation for sports and medicine, currently present, for example, in the MPU-9250 and ICM-20948 state of the art InvenSense MEMS sensors. The InvenSense MPU-6050 is older and limited (with only a 3-axis gyroscope and a 3-axis accelerometer, and no 3-axis magnetometer).

Why writing "6. Patents", if the developers of the new inertial sensor system, referenced in [20] of the manuscript, are not authors of the manuscript?

Author Response

All the authors thank kindly the reviewer for the constructive feedback.

Equation 3 and 4 seems not correct (Hamilton product, or simple product?)

This is indeed the Hamilton product (or quaternion product as defined in Dumas et al. 2004).

Using the quaternion product in this way is equivalent to multiplying by a rotation matrix in order to express the acceleration vector in another reference frame (in our case, the global frame computed during the calibration process). Once in the global frame, it is possible to subtract the gravitational acceleration (in a quaternion form) ; thus obtaining the linear acceleration in the global frame.

The regression and the squat detection algorithm are not described in detail in the manuscript.

The squat algorithm is detailed at line 194-196

The regression method is detailed at line 201-203

The magnetometer is an important sensor that should also be used in 3D body orientation for sports and medicine, currently present, for example, in the MPU-9250 and ICM-20948 state of the art InvenSense MEMS sensors. The InvenSense MPU-6050 is older and limited (with only a 3-axis gyroscope and a 3-axis accelerometer, and no 3-axis magnetometer).

The reviewer is right, and we already added a point in the discussion section saying the results could be improved with this extra sensor (line 303).

Nevertheless, the current IMU system has been designed to be used by physiotherapists in specific environments (e.g. hospital or clinics) where the magnetic disturbances can be important due to various equipment (MRI, elevators, power line…). The first approach was to focus on the motion capture precision without having to develop orientation algorithm with robustness to this type of disturbances.

Why writing "6. Patents", if the developers of the new inertial sensor system, referenced in [20] of the manuscript, are not authors of the manuscript?

The reviewer is right, this section was forgotten from the Sensors template and has been suppressed.

Reviewer 3 Report

5. 
   

   Manuscript# sensors-1477815

    

   Title: Three-steps validation of a new wireless inertial sensor system for embedded rehabilitation of low back pain

    

   General Comments

   The goal of this study was to validate a newly designed IMU system using a 2 step procedure: a sensor validation using a 6DOF hexapod platform with arbitrary motions and a functional validation on subjects performing squats. The authors claim good precision of the IMU system and that a placement on the middle or the proximal part of the thigh segment is to be preferred for squat motion. The manuscript is generally interesting. However, there are several major points that require revision.

    

   Major comments

   1. Title and entire manuscript. The authors claim that the sensor system will be relevant for low back pain. To justify this connection, outcomes must be clearly defined. Specifically, how do angular velocities and accelerations as well as orientation of the thigh relative to the pelvis during a squatting motion relate to low back pain? Please revise the entire manuscript to the fact that the validation of angular velocities and accelerations as well as orientation of the thigh relative to the pelvis during a squatting motion is tested without relating it to a specific disease. However, please also consider that the value of such system would be to assess clinically or functionally meaningful parameters assessed during relevant tasks.
   2. The manuscript needs copy-editing by a native speaker to make the text better readable.
   3. It is unclear why parameters computed during the hexapod experiment by the IMU were compared to those computed by the Vicons system. The latter two are both indirect measurements, yet the actual data should be available from the controllers of the hexapod. Comparing the IMU data to the hexapod input would be a true validation rather than a mere method comparison between two different indirect methods (IMU vs. optical system). This would have been a major strengths of this study compared to most of the literature.
   4. Additional details are required in the methods description and “raw” data should be provided. For instance, it is unclear if maximum accelerations and velocities or acceleration and velocity trajectories were included in the analysis.

    

   Specific comments

   5. Title: Delete “for embedded rehabilitation of low back pain”.
   6. Title: “new” – please specify what is new in this system as off the shelf IMU sensors were used (MPU6050)
   7. Abstract: Lines 10-12 and 15-17 – repetition
   8. Line 23 “…application for more than a decade.”
   9. Line 28: for -> from
   10. Lines 29, 30: suffer from -> have
   11. Line 32: typically between the lower rib margins and the buttock creases -> in the lumbar spine region
   12. Line 40: relied -> relies
   13. Line 40: complaint -> complaints
   14. Line 41: The trunk -> Trunk
   15. Line 42: delete “measuring”
   16. Line 50: experimentations -> experiments
   17. Line 51: method -> methods
   18. Line 55: like -> such as
   19. Line 59: placing -> location
   20. Line 60: sensors -> sensor
   21. Line 60-61: “… sensor placement for it is known to create high discrepancies in the field of kinematics results” – This statement is unclear.
   22. Line 61: regroup -> group
   23. Line 69-77 and entire manuscript: Please clearly define outcomes to be measured and to be validated with the “new” system.
   24. Line 82: Why did you select these subjects? Different numbers of female and male participants? Participation should be volunteers – was there a conflict of interest for participating?
   25. Table 1: Please define that mean +/- SD are shown. Weight -> body mass (weight is measured in N, body mass in kg). Please use SI units (m).
   26. Line 93: What was the rationale of performing unilateral measurements?
   27. Figure 1: What is the sensitivity of the system to magnetic fields?
   28. Line 111: “at respectively 20%, 50% and 80% of the maximum linear and angular velocity…” Please specify what the maximum linear and angular velocity is. Is it the range of the inertial sensors or values observed in human motion or values producible by the platform?
   29. Line 114: Please provide more details on the Vicon system: sampling frequency, model, place, country.
   30. Line 128: Please provide additional details on the task. How deep was the squat?
   31. Line 128: sessions -> trials
   32. Line 133: Please provide details on how the orientation, velocities and accelerations were computed. Did you use the manufacturer software?
   33. Line 139: is -> was
   34. Line 140: “…occasionally some packets were not transmitted.” What was the reason for this? Please describe the quality assurance procedures.
   35. Line 143-145: The camera data was resampled at 50Hz to allow for comparison to IMU data. Why were different cut-off frequencies then used for the two systems?
   36. Line 146-148: Lack of data synchronization – why was this not done? Please show data so the reader can understand what the time patterns were and what was used for further analysis.
   37. Line 152: to –> from
   38. Line 154: have to compute -> computed
   39. Line 164: With -> where… is
   40. Line 199: linear acceleration and angular velocities were used for sensor validation analysis – what is the relevance of these parameters in a clinical or functional context?
   41. Table 2: What data are shown here? Max values? Correlations between patterns/trajectories?
   42. Figures 4 to 6: Please label all axes – vertical axis: optical – IMU. Please use the same scale on the horizontal and vertical axes for the b) figures.
   43. Figure 7: Why were non-parametric statistics used? “orientation angles” – Are these maximum angles or time series of angles? Please specify.
   44. Figure 8: Are data for only one trial per participant shown here? This data would be better presented as scatter plot.
   45. Line 266: “…for the physiotherapist”. Please provide literature supporting the use of accelerations and velocities as parameters considered by clinicians and/or therapists.
   46. Entire manuscript: numbers – use dots as decimal separator instead of commata.

Author Response

Please find the answers to the Reviewers' comments in the enclosed file.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The manuscript has not been sufficiently improved to warrant publication. The magnetometer is an important sensor that should also be used in 3D body orientation for sports, rehabilitation and medicine, currently present, for example, in the MPU-9250 and ICM-20948 state of the art InvenSense MEMS sensors. The InvenSense MPU-6050 is older and limited (with only a 3-axis gyroscope and a 3-axis accelerometer, and no 3-axis magnetometer). The calibration of the magnetometer is also an important procedure.

Author Response

In the name of all the authors, we regret to inform you that we cannot answer the last remark of reviewer #2. I would like to take the time to elaborate why we are unable to answer this remark.

The reviewer says “the manuscript has not been sufficiently improved to warrant publication” and provides only one remark about the magnetometer. Knowing that this was already mentioned it (verbatim) in the first round of reviews, it is difficult to understand where the reviewer is heading by insisting on this very point.

All the authors were in agreement with this remark during the first round of reviews. We added a discussion on how the magnetometer could have improved the results and we explained why we decided to design the study without it regarding our future clinical applications (avoiding magnetic disturbances which are frequent in clinics and hospitals).

The reviewer says “The magnetometer is an important sensor” and we completely agree on that point, and as a matter of fact, we never state the contrary in the manuscript about magnetometer not being an important sensor.

The reviewer says “The calibration of the magnetometer is also an important procedure” and we also agree but do not see the importance of going into further details on that point within our manuscript. It would be out of topic as our device does not use a magnetometer.

It is obviously impossible for us to change the raw data by adding an MPU-9250, or any IMU with a 3-axis magnetometer, because it would require to go over the whole 180 trials.

As a conclusion, the authors cannot provide further answers to the remark of reviewer #2.

It seems that reviewer #2 ‘s only point is blocking the publication process of the manuscript. We believe it is incorrect to consider that all motion capture studies are wrong based on the fact that they do not use IMU with magnetometer. We also believe it would be unfair to block our manuscript based on that rationale.

Reviewer 3 Report

Manuscript# sensors-1477815

Title: Three-steps validation of a new wireless inertial sensor system for clinical application

General Comments

The goal of this study was to validate a newly designed IMU system using a 3-step procedure: a sensor validation using a 6DOF hexapod platform with arbitrary motions and a functional validation on subjects performing squats. The authors claim good precision of the IMU system and that a placement on the middle or the proximal part of the thigh segment is to be preferred for squat motion. The manuscript is generally interesting. The manuscript has greatly improved upon implementation of the reviewers’’ suggestions. However, several points require further revision.

Major comments

1. In their response to the reviewers, the authors provide a length justification why wearable sensors maybe useful. There is no question about that. However, when validating novel sensors or comparing these to other methods, it is critical to focus on the outcome or parameter of interest. Even though the authors have removed the focus on low back pain, the still discuss the sensors in the context of clinical applications. Please provide outcome parameters to be assessed and relevant or focus on the validation and comparison of the physical parameters velocity and acceleration.
2. The authors did not provide a clean version of the revised manuscript. From the tracked changes version of the manuscript there still seem to be some grammar and article mistakes that requires copy-editing.
3. Regarding my previous comment: “It is unclear why parameters computed during the hexapod experiment by the IMU were compared to those computed by the Vicons system. The latter two are both indirect measurements, yet the actual data should be available from the controllers of the hexapod. Comparing the IMU data to the hexapod input would be a true validation rather than a mere method comparison between two different indirect methods (IMU vs. optical system). This would have been a major strength of this study compared to most of the literature.”
   The authors answer this concern by stating that the geometrical coordinates of the hexapod center were not known at the time of the acquisition. This is a major shortcoming of the methodology. As pointed out previously, this would have been a true validation of the system that would have represented a very important contribution to the literature.
   The authors only performed unilateral assessments. Yet, clinically, assessing limb symmetry is essential.
   Please revise the entire manuscript to clarify that the authors present a method comparison and not a validation of the IMU system.
4. The authors still claim that their system is new, yet the sensors are commercially available. The new contribution would have been a validation or method comparison of clinically relevant parameters which were not provided.
5. Concerning previous comment 24: The authors state in their response that the research was performed during COVID lockdown, and that they had to work with the researchers on site which impacted their ability to respect sex parity. The fact that only lab personnel was included raises ethical concerns. Please emphasize that all participants participated voluntarily. In situations like this, one must ensure that students (dependents) will not feel obliged to participate and would not being willing to participate would not impact their studies or work towards degrees.
6. The authors specified that they tested the range of acceleration and angular rate producible by the hexapod. Please specify how these relate to the magnitude of accelerations and angular rates observed in patients during relevant tasks. For instance, if accelerations during relevant tasks are much lower than the accelerations produced by the hexapod, the data provided here would be meaningless regarding the accuracy of the system for clinical applications.
7. Thank you for including the raw data of the signals.
8. Figure 8: The figure included in the response to reviewers is not a scatter plot. In a scatter plot, one would present the estimated total number of squats for the IMU on the vertical axis and the estimated total number of squats for the Vicon system on the horizontal axis. One would easily see that both estimates agree very well with each other.
9. The clinical relevance of the accelerations and velocities is still unclear. The outcomes in the cited studies are not segment accelerations and velocities. Please clarify.
10. Figure 5 to 7. Units are still missing on the axes. Please also provide more detailed labels, e.g. vertical axis “IMU-Vicon”.
    These figure show the difference between Vicon and IMU. However, because the authors consider Vicon to be the ground truth, they should report the difference between IMU and Vicon.
    These figures show that the agreement between the two systems is better for low accelerations than for high accelerations. This result must be discussed regarding relevance for accelerations observed during clinically relevant tasks.

Author Response

Please see the attachments for the response to Reviewer #3 and cover letter to the academic editor

Author Response File: Author Response.pdf