Review Reports - The Technical Development of a Prototype Lower-Limb Therapy Device for Bed-Resting Users

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This work proposed a portable lower-limb therapy device, which supports trainings on leg flexion/extension, ankle plantarflexion/dorsiflexion, and leg pressing with active loading. It features a user interface with interactive exergames. Experiments with 12 able-bodied participants were carried out, and the results were evaluated via questionnaires.

The paper is well-written and easy to follow. The reviewer has positive impressions about the device, especially on its mechatronics design and form factor. However, as an initial study intended to demonstrate the capability of the device, this manuscript lacks the depth needed for a journal publication. And more critically, the reviewer believes the experimental design is flawed: instead of relying on subjective questionnaire results, significant efforts should be made on system evaluation and characterization. In its current form, only brief and cursory findings were included (L228-246), with plots

showing only representative results. In the reviewer’s opinion, without a thorough understanding of the capability of the system/device (e.g. frequency responses), it is premature to begin testing with human subjects. A few example questions to answer:

- Without “external” disturbances from human subjects, how well does the device track a certain trajectory (position and torque) at a certain frequency?

- Does the motors induce non-trivial frictions at higher speed? Is feedforward compensation needed?

- How do the placement of the force sensors affect measurement accuracy and stability, and subsequently the force tracking?

- Does other external factors, such as the firmness of bed, and the stiffness of the backboard, introduce excessive noises, and potentially affect its overall performance?

Other comments:

- L130-152 in Section 2.3 are standard implementations - consider trim or skip

- What metrics were evaluated with the RMSE listed in Eq. (1)?

- In the game demonstrated in Figure 6, how is the force being controlled?

-- Is it a constant resistive force designed to be applied throughout the sinusoidal movement?

-- Was the force applied bi-directional, or only during knee extension (i.e. when the user is “pushing down”)? If it's bi-directional, how is force measured during knee flexion?

Minor comments:

- L105: use consistent indexing: “1)” instead of “i)”

- L197: typo “30o”

Author Response

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

Comment 1: This work proposed a portable lower-limb therapy device, which supports trainings on leg flexion/extension, ankle plantarflexion/dorsiflexion, and leg pressing with active loading. It features a user interface with interactive exergames. Experiments with 12 able-bodied participants were carried out, and the results were evaluated via questionnaires.

Authors' response 1: We agree with the reviewer that only brief results of movement control were presented in our last manuscript. Based on the comments, the numerical results of movement control from all participants are presented in Table 2 and discussed in Lines 257-259, 270-272 in the updated manuscript.

The work described the development and evaluation of an in-bed lower-limb therapy device prototype. According to User-Centred Design process, two studies [1,2] were performed previously to collect the clinical requirements through interviews and focus group meetings. The first study [1] yielded the requirements of the device such as “Easy to operate, Portable, Can be fixed in bed, Quick to install in bed” as the first priorities, while the accurate force control and feedback for the patient as the attractive or optional features. In the second study [2], the desirable requirements were defined, as listed in the manuscript, as:

(1) it should be lightweight and compact for transport and storage,

(2) it should be easy to use, not only for clinical staff but also for patients without specialist supervision,

(3) it should provide therapeutic passive and active leg exercise,

(4) it should provide ankle dorsiflexion / plantarflexion, and

(5) it should provide an intuitive user interface with interactive exergames.

As a follow-up study of the previous two studies [1, 2], this work presented a compact device which can be operated easily, installed conveniently to provide leg and ankle movement with satisfactory accuracy. The developed system was evaluated to be acceptable, which met the requirements of the in-bed lower-limb therapy device. An emergency stop was implemented to secure the safety of test persons. The system was tested to be stable before the evaluation test on 12 participants. During the test, there was no instability-related issue. We believe the study results are interesting to be published to show the general functions, as the whole development of our device. The developed in-bed leg-press showed technical advancements in well-controlled movement, multi-modal training patterns, convenient operation, and interactive exergame. The work is innovative and fits the scope of the current special issue “Intelligent Systems, Robots and Devices for Healthcare and Rehabilitation".

From the perspective of research on accurate force control, we agree with the reviewer that it is interesting to investigate the force control strategy using frequency analysis. That is exactly what we are doing as the further development of the device. We developed a feedforward compensator using frequency response in our previous study [3]. Recently, we addressed rejection of speed-related disturbance to force control through an innovative method. This method was summarised in the attached document (Frequency_analysis_feedforward speed compensator.pdf), which will be improved and submitted to the journal Actuator soon. We have implemented the same method in the “Lower-limb Therapy Device”. Promising results have been obtained. As you can see from the attached document, this method involves many tests (open-loop, close-loop), identification of the system transfer function and velocity disturbances. These results are too much to be published in the current work.

Our paper publication plan is as follows: After the general performance of the in-bed lower-limb therapy device is published through the current work, we will publish our next paper titled: “An interactive game-based in-bed leg training system: control development and clinical usability evaluation”. This paper will present the development of an improved force control strategy using frequency-domain analysis with speed feedforward compensator. Furthermore, the clinical feasibility of the improved system will also be evaluated. We would like to ask you to review our future work if you are interested.

To address the comments from the reviewer, we highlighted the safety of the system through an emergency stop in Fig. 2(a), rewrote the discussion about the control section (Lines 351-356), future work (Lines 322-324), and highlight the strengths (Lines 357-359) and advancements (Lines 387-389) of the lower-limb therapy device.

References:

Rostetter, R.; Jenni, B.; Eggmann, S.; Meyer, JT.; Schmitt, K-U. Implementing an interprofessional user-centered design approach to develop a bedside leg exercise device. Technol Health Care. 2022,30(4):981-92.
Cerrito, A.; Fang, J.; Schertenleib, S.J.G.; Hunt, K.J.; Schmitt, K-U. First iteration of a User-Centered Design process to develop an in-bed leg press. Technology and Health Care. 2024, 32(6): 4637-4651
Fang, J.; Haldimann, M.; Marchal-Crespo, L.; Hunt, K.J. Development of an active cable-driven, force-controlled robotic system for walking rehabilitation. Frontiers in Neurorobotics. 2021,15(49).

Comment 2: A few example questions to answer:

- Without “external” disturbances from human subjects, how well does the device track a certain trajectory (position and torque) at a certain frequency?

Authors' response 2: The frequency analysis will be performed in the future. In the current study, the position control of the linear axis was performed only at a fixed frequency (a cycle time of 6 s). The movement control accuracies are presented in Table 2 of the updated manuscript, which are:

The position control of the linear axis was performed with a target sinusoidal curve of a magnitude of 250 mm with a cycle time of 6 s. The position tracking error was 8.91 ± 3.49 mm.

The position control of the ankle joint was performed with a target sinusoidal curve of a magnitude of 26^o with a cycle time of 6 s. The position tracking error was 1.62 ± 0.87^o.

The force control of the linear axis was performed with a target force of 50 N, while the participants voluntary moved the leg following a sinusoidal curve of a magnitude of about 250 mm with a cycle time of 6 s. The force tracking error was 2.52 ± 0.72 N.

Comment 3: - Does the motors induce non-trivial frictions at higher speed? Is feedforward compensation needed?

Authors' response 3: Yes, when the motor is moving, the induced friction cannot be ignored. The PID control strategy partly compensated the influence of friction. This can be seen in Fig. 8. For a constant target force, the required torque should be constant at static resistive control, i.e. without any linear movement. However, during the movement, as shown in Fig.8b, a varying torque curve was required. The varying torque partly showed the compensation of the friction induced during movement.

When the movement speed is higher, then the friction is believed to be higher, too. The PID controller will automatically produce the required torque at a higher speed. But integration of a speed compensation is desirable. This is exactly what we are investigating at the moment. The results will be presented in the follow-up paper.

In the current work, the movement speed was low. The PID control strategy produced satisfactory movement accuracy (Table 2).

To address this comment, we added in Lines 350-357: “Additionally, test results indicated that force control accuracy was influenced by the voluntary movement of the foot platform. Velocity disturbances are a common issue in force control [21]. This test implemented force at a fixed speed (metronome of 20 BPM). If the metronome beat are higher, the test person will physically move their feet at a higher speed, resulting in a higher influence of the movement speed in the force control accuracy. The force control strategy of the lower-limb therapy device can be improved by integrating a feedforward velocity compensator. These control strategies will be explored in future research.”

Comment 4: - How do the placement of the force sensors affect measurement accuracy and stability, and subsequently the force tracking?

Authors' response 4: The placement of the force sensors affects measurement accuracy and stability. If we fix the force sensor directly on the motor shaft where the foot platform unit is fixed, the force measurement includes the force that is required to move the heavy foot platform, apart from the active force from the participant. Therefore, we fixed the sensor just below the foot plate, where the weight of the foot platform was supported by the mechanical base. In addition, the type of force sensor affects the measurement stability. We previously used a three-dimensional (3-D) force sensor. The force measurement in the other two axes influenced the force control algorithms. Therefore, in the current study, one-dimensional (1-D) force sensor was used.

To address this comment, we made changes in the updated manuscript as follows:

Fig. 2b with placement of the fore sensors was added. Furthermore, in Lines 104-110 of the updated manuscript, we wrote: “Two force sensors (one-dimensional pull/press sensor; resolution 0.12 N; range of -500 N to 500 N; Bengbu Sensor Company, China) were mounted under each foot plate (Fig.1 and Fig. 2b) to measure the active force from the user. The force sensor was calibrated. The placement of the force sensors affected measurement accuracy and stability. The force sensor was fixed directly below the foot plate, removing the influence from the mechanical components on the force measurement.”

Comment 5: - Does other external factors, such as the firmness of bed, and the stiffness of the backboard, introduce excessive noises, and potentially affect its overall performance?

Authors' response 5: Yes, the external factors, such as the firmness of bed, and the stiffness of the backboard, introduce excessive noises, and potentially affect its overall performance. To prevent these influences, we fixed the bed by four brakes. We set the target force to be low (50 N), so that the participant did not need to move the trunk or a firm trunk support so as to perform the resistive training.

To address this comment, we added in Line 211 “The therapy bed was stablised by four brakes.”

Comment 6: - L130-152 in Section 2.3 are standard implementations - consider trim or skip.

Authors’ response 6: This part provided important method information of how we developed the overall programmes, established communication between raspberry Pi and the movement controller MiniMACS, and implemented the control algorithms. This is also important for those who want to achieve the similar function. But as suggested by the reviewer, we trimmed this section in the updated manuscript, as shown in Lines 152-170.

Comment 7: - What metrics were evaluated with the RMSE listed in Eq. (1)?

Authors’ response 7: The metrics presented in Table 2 in the updated manuscript were evaluated using this equation, which are the accuracy of the force control in the active mode, and the position control (the linear position and ankle joint rotation) in the passive mode. This was stated in Lines 248-252 in the updated manuscript.

Comment 8: - In the game demonstrated in Figure 6, how is the force being controlled?

Authors’ response 8: The force control strategy is the same as the active mode (described in the second paragraph of Section 2.2), where a constant force was set as the target, and the PID control strategy (Fig. 3b) was implemented.

To address this comment, we made following changes in the updated manuscript:

In Lines 192-193: “During training with the exergames, the force control strategy shown in Fig. (3b) was implemented on both legs.”

Comment 9: -- Is it a constant resistive force designed to be applied throughout the sinusoidal movement?

Authors’ response 9: Yes. A constant resistive force was applied. The participant controlled the trajectory voluntarily.

To address this comment, we made following changes in the updated manuscript:

In Lines 224-225: “During the active and game sessions, a target force of 50 N was defined for each foot.”

Comment 10: - Was the force applied bi-directional, or only during knee extension (i.e. when the user is “pushing down”)? If it's bi-directional, how is force measured during knee flexion?

Authors’ response 10: Only resistive force was trained, which means that the participant pressed the foot platform while knee extension. The leg pressing information was provided all through the manuscript as follows:

Line 19: “leg pressing”

Line 115: “3) active leg pressing”

Line 138: “…produce active loading during leg pressing.”

Line 189-190: “The moving object could be controlled by pressing the right or left foot plates with a certain displacement.”

Lines 219-221: “In active session, the participant pressed against the foot platforms in order to stretch both legs reciprocally.”

Lines 222-223: “In game session, the participant pressed against the foot platforms and moved both feet”

Comment 11: Minor comments:

- L105: use consistent indexing: “1)” instead of “i)”

- L197: typo “30o”

Authors’ response 11: This was corrected in the updated manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript addresses an important topic in rehabilitation engineering and offers a novel solution for the early mobilization of bed-resting patients. Integrating interactive games adds an engaging dimension to therapy and is highly relevant to current trends in patient-centered care. The technical aspects of the device are well-explained, and the preliminary results are promising. However, the study’s evaluation is limited to able-bodied participants, which may not fully validate its effectiveness for the target patient population. The main focus and hypothesis need to be clearly articulated in the introduction - whether it aims to develop a lower-limb therapy system or evaluate its effectiveness based on the proposed user-centered design and interactive games.

Section 2.1 lacks more details about the technical characteristics of the hardware components used in the device, such as the type of force sensors used (only the max range is mentioned, but no resolution, type of force sensor, model, etc.). How was the ankle rotation measured? How was the range of linear movement ensured? Were any (if so, what kind of) safety measures implemented in the device design? How was the force and position control accuracy validated? Was it solely based on the readings from the embedded sensors? If so, was any calibration performed to ensure the accuracy of these sensors?

Section 2.5: Subjects were instructed to move to the beat of the metronome. What tempo was used, and was it the same for all participants? How the change in tempo would affect the exercise?

Section 3: based on what performance characteristics were the parameters for the PID controller selected? This kind of technical article in the Actuators journal would highly benefit if more details on the control algorithm development were provided. A visual reference for how the device or participant’s limbs were oriented during the tests (Figure 8 or 7).

How were participant responses to the questionnaire analyzed? Was any statistical testing performed to assess trends or significance?

The conclusions would benefit from greater specificity about the technical advancements and their potential clinical impact.

Formatting errors:

Line 197 - "300 mm displacement and 30o rotation in the ankle joint" - should be correctly formatted with the degree symbol.

Author Response

Comment 1: The manuscript addresses an important topic in rehabilitation engineering and offers a novel solution for the early mobilization of bed-resting patients. Integrating interactive games adds an engaging dimension to therapy and is highly relevant to current trends in patient-centered care. The technical aspects of the device are well-explained, and the preliminary results are promising. However, the study’s evaluation is limited to able-bodied participants, which may not fully validate its effectiveness for the target patient population. The main focus and hypothesis need to be clearly articulated in the introduction - whether it aims to develop a lower-limb therapy system or evaluate its effectiveness based on the proposed user-centered design and interactive games.

Authors’ response 1: The aim of this work was to technically develop and to evaluate the usability of an in-bed lower-limb therapy device prototype that provides various training patterns for the foot and leg with an intuitive user interface and interactive exergame. The work here presented a preliminary assessment of the usability in able-bodied participants. Follow-up clinical studies will be performed on the end-users (patients and therapists) to evaluate the clinical feasibility and to investigate the effectiveness of the developed system.

To address this comment, we made following changes in the updated manuscript:

In Lines 88-93: “The aim of this work was to develop and evaluate the usability of a prototype in-bed lower-limb therapy device that offers various training patterns for the foot and leg, featuring an intuitive user interface and interactive exergame. This work presented a preliminary assessment of usability in able-bodied participants. Follow-up clinical studies will be conducted on end-users (patients and therapists) to evaluate clinical feasibility and investigate the effectiveness of the developed system.”

In Lines 315-316: “This work developed and evaluated the usability of a prototype in-bed lower-limb therapy device that …”

In Lines 323-325: “Future work includes improving the control strategy, increasing the reaction speed of the computer game and conducting clinical studies for evaluation of feasibility and effectiveness of the developed system.”

Comment 2: Section 2.1 lacks more details about the technical characteristics of the hardware components used in the device, such as the type of force sensors used (only the max range is mentioned, but no resolution, type of force sensor, model, etc.). How was the ankle rotation measured? How was the range of linear movement ensured? Were any (if so, what kind of) safety measures implemented in the device design? How was the force and position control accuracy validated? Was it solely based on the readings from the embedded sensors? If so, was any calibration performed to ensure the accuracy of these sensors?

Authors’ response 2: The type of the force sensor is one-dimensional pull/press force sensor from Bengbu Sensor Company, China; resolution 0.12 N; the force range of -500 N to 500 N.

The ankle rotation was measured via the encoder on the motor for the ankle rotation. The linear movement was measured via the encoder on the motor for the linear movement. There were limit switches on both ends of the linear movement.

There are safety measures implemented in the device design. An emergency stop was implemented (See Fig. 2a).

The force and position control accuracies were calculated using Eq. (2). The accuracy was solely validated based on the reading from the force sensor and the encoder on the motors. Yes, the force and encoder sensors were calibrated in advance.

To address this comment, we made following changes in the updated manuscript:

In Lines 104-107: Two force sensors (one-dimensional pull/press sensor; resolution 0.12 N; range of -500 N to 500 N; Bengbu Sensor Company, China) were mounted under each foot plate (Fig.1 and Fig. 2b) to measure the active force from the user. The force sensor was calibrated.

In Lines 134-136: “The actual linear movement and the ankle rotation was measured via encoder on the motors. These position measurements were calibrated. There were limit switches on both ends of the linear movement.”

Fig. 2a was added to show the emergency stop.

Comment 3: Section 2.5: Subjects were instructed to move to the beat of the metronome. What tempo was used, and was it the same for all participants? How the change in tempo would affect the exercise?

Authors’ response 3: The beat of the metronome was set to be 20 beats per minute. It was the same for all participants. When the tempo is higher, then the test person will physically move their feet at a higher speed, resulting in a higher influence of the movement speed in the force control accuracy.

To address this comment, we made following changes in the updated manuscript:

In Lines 221-222: “They moved to the beat of a metronome, which was set to be 20 beats per minute for all participants.”

In Lines 352-355: “This test implemented force at a fixed speed (metronome of 20 BPM). If the metronome beat is higher, the test person will physically move their feet at a higher speed, resulting in a higher influence of the movement speed in the force control accuracy”

Comment 4: Section 3: based on what performance characteristics were the parameters for the PID controller selected? This kind of technical article in the Actuators journal would highly benefit if more details on the control algorithm development were provided. A visual reference for how the device or participant’s limbs were oriented during the tests (Figure 8 or 7).

Authors’ response 4: The PID loop parameters (proportional, integral and derivative gains) were tuned manually by trial and error. A step response was performed, and the performance characteristics of rising time shorter than 0.5s and the errors smaller than 5% was used.

To address this comment, we made following changes in the updated manuscript:

The PID control algorithms were presented in Lines 130-134:

In Lines 145-150: we wrote “Using the Scope function in ApossIDE, the parameters of the three PID controllers were tuned based on the experimental trial and error method [14]. The integral and derivative gains k_I and k_D were initially set as zero, while the proportional gain k_P was gradually increased so that the actual position or force tracked the step response with acceptable performance (rising time shorter than 0.5 s, tracking error smaller than 5%). Then the integral and derivative gains k_I and k_D were slightly tuned to refine the movement.”

Regarding the comments of “A visual reference for how the device or participant’s limbs were oriented during the tests”, we made following changes in the updated manuscript:

Two pictures were included in the updated manuscript (Fig. 2b and Fig. 6c) to show a visual reference for how the device and participant’s limbs were oriented during the tests. Furthermore, a video was uploaded to show the movement of the test, as described in Lines 225-226: “A video of the game session was attached to show how the device and the participant’s limbs were oriented during the tests.”

Comment 5: How were participant responses to the questionnaire analyzed? Was any statistical testing performed to assess trends or significance?

Authors’ response 5: The participants responses were analysed using a modified System Usability Scale [1]: 13 questions were asked. The answers to the questions were analysed on a scale of 5, where “1” to “5” means “strongly disagree”, “disagree”, “neutral”, “agree” and “strongly agree”, respectively.

As this is a qualitative analysis, not a hypothesis-driven study, therefore, no statistical analysis was required. For the qualitative analysis, the results and standard deviation (mean+SD) were analysed in Tables 3-4, as recommended in study [1].

Reference:

[1]. Bangor A, Kortum P, Miller J. Determining what individual SUS scores mean: Adding an adjective rating scale. J Usability Stud. 2009;4:114-23.

To address this comment, we made following changes in the updated manuscript:

In Lines 227-228: “13 questions were prepared based on a modified System Usability Scale [16].”

In Line 245: “A qualitative analysis was performed using the evaluation method proposed in [16].”

In Lines 295-301: “A System Usability Scale score of 70%, or 3.5 out of 5 points, indicates good usability, while 4.0 out of 5 points means excellent usability [16]. Summarising the questionnaire results of the 12 participants (Tables 3-4), a mean score higher than 4.0 points was obtained in 5 out of 7 questions regarding the mechanics and movement control yield, and 4 out of 6 questions related to the user interface and computer game. Therefore, the usability of the in-bed lower-limb therapy device was rated as excellent apart from Q1.3, Q1.5 and Q2.3.”

Comment 6: The conclusions would benefit from greater specificity about the technical advancements and their potential clinical impact.

Authors’ response 6: We rewrote the conclusions, as seen in Line 383-392 of the updated manuscript.

Comment 7: Formatting errors:

Line 197 - "300 mm displacement and 30o rotation in the ankle joint" - should be correctly formatted with the degree symbol.

Authors’ response 7: This was corrected in the updated manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for submitting your paper. Although I believe that the idea of providing robotic therapy devices for in-bed exercises is an interesting and important idea, your device still requires some improvements.

I believe that if you want to go ahead and publish your work to date you must add the words "A Prototype" to the title and rewrite your discussion and conclusion as I have indicated in the manuscript.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

There are some minor details re the English used in this manuscript which I have commented upon.

Author Response

Comment 1: Thank you for submitting your paper. Although I believe that the idea of providing robotic therapy devices for in-bed exercises is an interesting and important idea, your device still requires some improvements. There are some minor details re the English used in this manuscript which I have commented upon

Authors’ response 1: Thank you for the comments. We updated the manuscript as you recommended. We rewrote the discussion and conclusions. The future work was defined to improve the device. The all manuscript was thoroughly updated as you recommended. The detailed response can be found in the attached PDF “Updated_actuators-3373984 - for Reviewer3_final”.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

This reviewer does not have further comments.

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for your responses to my comments. You have adequately addressed all my concerns.