TRIPOD—A Treadmill Walking Dataset with IMU, Pressure-Distribution and Photoelectric Data for Gait Analysis

Round 1

Reviewer 1 Report

I would like to thank the authors for their changes made to the manuscript. In my opinion, the current manuscript more clearly explains what this manuscript contributes to the current body of evidence. The majority of my questions have been addressed. I have some additional comments I'd like the authors to consider:

• The authors choose to remove outliers. Can the authors elaborate to a larger extent why they choose to do so? Having a reference system allows the removal of these data points. However, without a reference system this may be harder. If this is the case, the results of this trial are not necessarily translatable to a situation where there is no reference system. Therefore, I suggest the authors to justify this choice better.
• In my opinion, the part under Figure 9 (line 298-302) should not be in the result section, but should be part of the discussion section.
• In my opinion, the discussion section is too short and is not focusing on the themes it should focus on. I think this is explained that part of the result section should be in the discussion section. I would suggest the authors to report the results without interpretation in the result section and shift all parts where they interpret or explain the results to the discussion section. Furthermore, I miss a clear conclusion in the discussion section. For instance, is the IMU system good enough for the outcomes that the authors report?

Author Response

Please see the attached

Author Response File: Author Response.pdf

Reviewer 2 Report

This publication presents a dataset for IMU-based gait analysis that consists of 15 healthy young adults walking on a treadmill and includes IMU measurements and ground truth data from a Zebris instrumented treadmill and an OptoGait system. The authors also present a Python-based data processing pipeline.

The manuscript is well written and the experimental procedure and preprocessing of the data are clearly described. There is a need for openly available datasets in the field of IMU-based gait analysis and the TRIPOD dataset certainly provides a valuable contribution.

The resubmitted manuscript contains significant improvements, and the authors adequately address all major points that were previously raised.

While it is unfortunate that the data is only available upon request, this is now clearly indicated in the abstract and the manuscript. However, the explanation given by the authors regarding the ethics approval is not available to the reader. It is also not made clear how the information provided in the request form is used. I suggest two changes to the request form:

1. Add one sentence that explains why providing data is necessary.

2. Add one sentence that explains how the provided information is stored and for what purposes it is used.

Author Response

Please see the attached

Author Response File: Author Response.pdf

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 publication presents a dataset for IMU-based gait analysis that consists of 15 healthy young adults walking on a treadmill and includes IMU measurements and ground truth data from a Zebris instrumented treadmill and an OptoGait system. The authors also present a Python-based data processing pipeline.

The manuscript is well written and the experimental procedure and preprocessing of the data is clearly described. There is a need for openly available datasets in the field of IMU-based gait analysis and the TRIPOD dataset certainly provides a valuable contribution.

A major concern is that the dataset is available only after providing extensive personal details and information regarding the planned purpose and after the authors accepted the request. Not having the data publicly available significantly reduces the value of this contribution to the scientific community. Furthermore, the nature of the data is not very sensitive (e.g. it does not contain patient data) and no explanation is given for why access is limited. At the very least, the authors must refrain from using the term "publicly available" in the abstract and check the manuscript for similar misleading statements.

The manuscript still offers room for improvement in the following places:

- The term "Optical Data" used in the title is very generic. While not inaccurate, many researchers would expect marker-based optical motion capture data and expect to be able to use this data to analyze the full 3D gait kinematics. It might therefore be worth considering changing the title in a way that clarifies the nature of the optical data. Also in the text (e.g. "optical gait analysis system" in line 67) more precision regarding this will be helpful for the reader.

- The authors state that the Zebris software does not provide any parameters except for stance time (line 98). This statement is not accurate as a wide range of parameters (various gait phase durations, stride length, cadence, ...) can be exported to csv files as mean+standard deviation for a sequence of steps. Being able to use metrics calculated by the system manufacturer (as opposed to metrics derived from the raw data with a custom algorithm) would increase the value of the TRIPOD dataset. The authors should therefore find a way to export this data and add it to the dataset.

- The description of the sensor attachment does not include information about the intrinsic sensor axes. Coordinate systems should e.g. be added to Figure 1 to show how the sensors were attached.

- The description of the data processing pipeline lacks detail about the scope and limits of the software. The authors do not state
    - which gait parameters *can* be calculated using the pipeline and in which way this is extensible
    - which parameters *are* currently calculated from the three measurement systems
    - which kind of evaluation is performed and which output data is produced.

- For the processing of the Zebris raw data, a detailed description of the implemented algorithm should be added to the manuscript.

- In the evaluation, the authors only show results for stride length and stride time in Figure 4. Stance and swing duration are excluded as the measurement systems show a systematic difference. However, in order to later use this ground truth data for assessing the validity of IMU-based methods, analyzing this difference is crucial. Figures showing those results must be included and discussed in-depth. The same applies to the comparison of stance and swing durations obtained from the IMU measurements.

- The authors should carefully check that the provided code runs out-of-the-box. In a quick test, two small problems were noticed:
    - import not working in src/pipeline/reference_loader.py ("no module named src")
    - example data files are not found due to capitalization ("OptoGait" vs "Optogait")
Furthermore, the README.md file should include some details of what output is to be expected when running the provided example script.

Language/spelling and other details:
- the exact meaning of the name TRIPOD is not explained in the manuscript
- lines 27 and 33: the word "safety" should probably be used instead of "security"
- line 105: comma missing after "additionally"
- line 121: "self-selected"
- line 188: extra hyphen in "estimation-"
- line 218: extra hyphen in "stance-"
- Bland-Altman plots typically show the mean of both measurements on the x axis
- Figure A1: label is cut off

Reviewer 2 Report

The authors collected and present a dataset of treadmill walking as measured by three different measurement systems. A similar and more extensive database is already available, which limits the novelty of this paper. Furthermore, the methods section lacks important details.

 

Introduction section

The authors state that to the best of their knowledge no database and software framework exists of IMU data collected on a treadmill. I would like to point the authors towards the work done by Loose and Bolmgren in which data of 108 non-impaired individuals are collected and made publicly available: http://gaitanalysis.th-brandenburg.de/. I would suggest the authors to include this reference in the introduction section and explain how their database is different from the existing database and what the contribution the presented database is to the existing body of knowledge.

 

Methods

• In Table 1 the authors present the preferred walking speed. Is this the preferred walking speed as established on the treadmill?
• Could the authors elaborate on the calibration procedures that were used on the IMUs?
• Were data filtered? If so, please provide filter characteristics.
• Could the authors explain how stride and step length were calculated using the IMUs?