CTR056048 Supplier Grant Application_Lee.pdf

TABLE OF CONTENTS

General Information 1

W-9 2

Acknowledgement Statement of Acceptable Terms and Conditions 3

Review Purposes 4

Project Abstract 5

Detailed Project Description

Goals and Objectives 6

Background Information and Literature Review 6

Preliminary Work 7

Experimental Methodology 8

Data Analysis 11

Timeline 11

Evaluation 11

Project Resources 12

References 14

Biographical Sketches 18

Hyunglae Lee 19

Megan Eikenberry 24

Carolyn Kinney 26

Letters of Support/ Mentor Support 29

Budget Information 33

Certifications and Assurances 37

Attachment C 41

Attachment D 42

Attachment E 43

Attachment F 45

APPLICATION PACKET

ATTACHMENT B

RFGA NO. BPM000338

General Information

Information about the Applicant organization or individual and the person to contact for any questions during the review process should be clearly listed.

1. Provide the following contact information for the person who is responsible for answering questions related to this Application:

Name Nicole White

Title Grant + Contract Specialist

Mailing Address PO Box 876011, Tempe, AZ 85287-6011

Phone Number 480727-3412

Email Address Asu.awards@asu.edu

2. Legal name and address of Applicant:

Name AZ Board of Regents on behalf of Arizona State University

Mailing Address PO Box 876011, Tempe, AZ 85287-6011

W-9 Attached Attached

3. Name of Principal Investigator:

Name Hyunglae Lee

Email Address hyunglae.lee@asu.edu

Date of terminal degree or completion of medical residency

May, 2013

4. Title of Project

Title Patient-adaptive robotic balance training for lower-extremity stroke rehabilitation

5. Total Fund Request for Project Budget Period (identify dollar amount per year):

Year 1 $75,000

Year 2 $75,000

Year 3 $75,000

Total Requested $225,000

PI Hyunglae Lee, General Information, BPM000338, 1

Form W•9 (Rev. October 2018) Department of the Treasury

Request for Taxpayer Identification Number and Certification

Give Form to the requester. Do not send to the IRS.

Internal Revenue Service .- Go to www.irs.gov/FormW9 for instructions and the latest information.

1 Name (as shown on your income tax return). Name is required on this line; do not leave this line blank.

Arizona State Universitv 2 Business name/disregarded entity name, if different from above

~ 3 Check appropriate box for federal tax classification of the person whose name is entered on line 1 . Check only one of the 4 Exemptions (codes apply only to certain entities, not individuals; see instructions on page 3):

gi following seven boxes.

D.

§ 0 Individual/sole proprietor or [] C Corporation [] S Corporat ion [] Partnership [] Trust/estate ai ~ single-membet LLC Exempt payee code (it any) __ 3 __

0. 0 }!- t O Limited liability company. Enter the tax classification (C=C corporation, S=S corporation, P=Partnership) ... ___ _ c5 2 Note: Check the appropriate box in the line above for the tax classification of the single-member owner. Do not check .S ~ LLC if the LLC is classified as a single-member LLC that is disregarded from the owner unless the owner of the LLC is

Exemption from FATCA reporting

,. another LLC that is not disregarded from the owner for U.S. federal tax purposes. Otherwise, a single-member LLC that

o. ~ is disregarded from the owner should check the appropriate box for the tax classification of its owner.

code (if any) C

·0 0 Other (see instructions)... Government ~ ~5-'=A~d~d~re~s=s~(n=u=m~b~e=r.~s~tr~e=et~.~an~d- a_p_t-.o- r_s_u-it_e_n_o_.) _S_ee- in-st-ru_c_t-io-ns-. _ _::::.:::..::..:::.:..:.:.:..:.,.::.:..:..:: _____ -.-~R-e-q-ue-s-te-r~'s_n_a_m_e_aLn-d~a-d~d-re_s_s~(o-p~ti~o-na~I) _____ _

(Applies to accounts mBlntained outside the U.S.)

Q) cJ5 PO Box 876011 6 City, state, and ZIP code

Tempe,_Az ... 85287-601 .. 1 _ _ __ . _ __ .. _______________ _____________ .. __ .. ___________________ _ 7 List account number(s) here (optional)

•!?:'T·•• Taxpayer Identification Number (TIN) Enter your TIN in_ the appmpri'."te box. The TIN provided must. match t~e name given on line 1 to avoid backup withholding. For 1nd1v1duals, this 1s generally your social security number (SSN). However, for a resident alien, sole proprietor, or disregarded entity, see the instructions for Part I, later. For other entities, it is your employer identification number (EIN). If you do not have a number, see How to get a TIN, later.

I Social security number

[Ill -[D -I J __ J __ J Note: If the account is in more than one name, see the instructions for line 1. Also see What Name and Number To Give the Requester for guidelines on whose number to enter. i

6 19

8 6 - 0 6

Certification Under penalties of perjury, I certify that:

1. The number shown on this form is my correct taxpayer identification number (or I am waiting for a number to be issued to me); and

2. I am not subject to backup withholding because: (a) I am exempt from backup withholding, or (b) I have not been notified by the Internal Revenue

Service (IRS) that I am subject to backup withholding as a result of a failure to report all interest or dividends, or (c) the IRS has notified me that I am no longer subject to backup withholding; and

3. I am a U.S. citizen or other U.S. person (defined below); and

4. The FATCA code(s) entered on this form (if any) indicating that I am exempt from FATCA reporting is correct.

Certification instructions. You must cross out item 2 above if you have been notified by the IRS that you are currently subject to backup withholding because you have failed to report all interest and dividends on your tax return. For real estate transactions, item 2 does not apply. For mortgage interest paid, acquisition or abandonment of secured property, cancellation of debt, contributions to an individual retirement arrangement (IRA), and generally, payments other than interest and · · u are·not required to sign the ce1titication, but you must provide your correct TIN. See the instructions for Part 11, later.

Sign Here

General Instructions Section references are to the Internal Revenue Code unless otherwise noted.

Future developments. For the latest information about developments related to Form W-9 and its instructions, such as legislation enacted after they were published, go to www.irs.gov/FormW9.

Purpose of Form An individual or entity (Form W-9 requester) who is required to file an information return with the IRS must obtain your correct taxpayer identification number (TIN) which may be your social security number (SSN), individual taxpayer identification number (ITIN). adoption taxpayer identification number (ATIN), or employer identification number (EIN), to report on an information return the amount paid to you, or other amount reportable on an information return. Examples of information returns include, but are not limited to, the following.

• Form 1099-INT (interest earned or paid)

Cat. No. 10231X

Date ... \Jll/l

• Form 1099-DIV (dividends, including those from stocks or mutual funds)

• Form 1099-MISC (various types of income, prizes, awards, or gross proceeds)

• Form 1099-8 (stock or mutual fund sales and certain other transactions by brokers)

• Form 1 099-S (proceeds from real estate transactions)

• Form 1099-K (merchant card and third party network transactions)

• Form 1098 (home mortgage interest), 1098-E (student loan interest), 1098-T (tuition)

• Form 1099-C (canceled debt)

• Forrn ·1099-A (acquisition or abandonment of secured property)

Use Form W-9 only if you are a U.S. person (including a resident alien), to provide your correct TIN.

If you do not return Form W-9 to the requester with a TIN, you might be subject to backup withholding. See What is backup withholding, later.

Form W-9 (Rev. 10·2018)

PI Hyunglae Lee, W-9, BPM000338, 2

ACKNOWLEDGEMENT STATEMENT OF ACCEPTANCE OF TERMS AND CONDITIONS

Vendor accepts the terms & conditions with the following exceptions:

1) Section 11.5 "Advertising, Publishing and Promotion": ASU requests that the word "Publishing" be stricken as it is in conflict with Section 16. "Publications", the first requires prior written approval and the second does not. ASU can agree to Section 16, but cannot agree to Section 11.5 as long as the word "Publishing" remains.

2) Section 12.4 “Gratuities”: ASU as an entity of the State is unable under the State Constitution and Law to agree to penalty recovery as they would be funds not allocated by the legislature.

PI Hyunglae Lee, Acknowledgement Statement, BPM000338, 3

RFGA NO. BPM000338

REVIEW PURPOSES

1. What field would you like your grant reviewed under? Aging and Disease: Chronic Diseases and Disorders Affecting the Elderly

1.1. If “Other,” please specify:

Click here to enter text.

2. What is the specialized area of proposed research? Click here to enter text.

Stroke rehabilitation

3. Does your proposed research/project include human subjects? Yes

3.1. Is your use of human subjects research exempt? No

3.1.1. If yes, please describe:

4. Is your research an NIH defined clinical trial? Yes

4.1. If yes, please describe:

A new 5-week robotic balance training program will be tested on 20 stroke survivors to determine the effectiveness of intervention on improving paretic ankle motor control, postural balance and locomotion.

The training program consists of 10 sessions (2 sessions/week; 1 hour/session) and its effectiveness will be evaluated by performing assessments using both quantitative outcome measures and popular clinical measures at 3 different time points: pre-training (baseline), post-training (5-week) and follow-up (3-month).

5. Does the research involve Vertebrate Animals? No

5.1. If yes, please attach IACUC Approval Animal Welfare Assurance Number.

PI Hyunglae Lee, Review Purposes, BPM000338, 4

RFGA NO. BPM000338

PROJECT ABSTRACT

All Applicants are required to submit an abstract of the proposed activity suitable for publication. The abstract should be informative to other persons working in the same or related fields and, insofar as possible, understandable to a scientifically or technically literate lay reader. The abstract shall not include any proprietary or sensitive business information as it may be made available to the public.

The abstract shall be within 1,400 characters in length.

Stroke is one of the leading and most lasting causes of long-term disabilities in Arizona. Among possible sensorimotor consequences of a stroke, impaired control of postural balance is a major cause of chronic disability that heavily impacts the quality of life of stroke survivors, by making daily activities more challenging and increasing the risk of falls. With an overarching goal to improve the effectiveness of postural balance control in stroke rehabilitation beyond current best clinical practices that often lead to motor recovery plateau (diminished capacity for additional motor improvement), we propose an innovative robotic balance training approach that effectively identifies altered ankle mechanics following stroke and integrate this advanced knowledge into the design and execution of postural balance training. We expect that the 5-week robotic balance training program based on assist-as-needed and patient-adaptive control schemes will not only benefit the overall lower extremity function including postural balance and walking in stroke survivors, but also further lead to secondary benefits of improving the quality of daily activities and reducing relevant healthcare cost. Importantly, the proposed approach utilizing a patent-pending robotic balance platform will be filed for a new patent, which will impact the local economy by adding a new biomedical patent in Arizona.

PI Hyunglae Lee, Project Abstract, BPM000338, 5

RFGA NO. BPM000338

DETAILED PROJECT DESCRIPTION (PROJECT ACTIVITIES)

1. Goals and Objectives Current best clinical practices for postural balance control in stroke rehabilitation often lead to motor recovery plateau, i.e., diminished additional motor improvement in the chronic phase.1,2 With an overarching goal to improve beyond current best clinical practices, we propose an innovative robotic balance training approach that identifies altered ankle mechanics following stroke and provides patient-adaptive intervention.

Aim 1: Determine how stroke impacts 2D viscoelastic ankle mechanics during upright postural balance. Our objective is to quantify two-dimensional (2D) viscoelastic ankle mechanics in both the sagittal and frontal planes while stroke survivors are maintaining upright postural balance and to compare these mechanics to the baseline of age-matched, unimpaired controls.

Aim 2: Develop a patient-adaptive robotic balance training program targeting altered ankle mechanics.

Our objective is to develop a robotic balance training program to correct the altered 2D viscoelastic ankle mechanics in stroke survivors toward unimpaired baseline. We will adopt an assist-as-needed control strategy to promote patient engagement and adjust the amount of robotic assistance in a patient-adaptive manner.

Aim 3: Assess the effectiveness of the robotic balance training program. Our objective is to evaluate the effectiveness of the proposed balance training program for correcting altered ankle mechanics toward the baseline and improving lower extremity function in stroke survivors. Assessments using both quantitative and clinical outcome measures will be conducted at pre-training, post-training (5-week) and follow-up (3-month).

Hypothesis 1: Viscoelastic ankle mechanics in stroke survivors are significantly different from those in age-matched, unimpaired controls. The rationale for this hypothesis is based on our preliminary results demonstrating significantly altered ankle stiffness (elasticity) and damping (viscosity) in chronic stroke survivors during upright postural balance, which is likely due to weakness, spasticity and contractures.3 Hypothesis 2: The proposed 5-week, patient-adaptive, robotic training will significantly change altered viscoelastic ankle mechanics toward the unimpaired baseline in both the sagittal and frontal planes. The rationale for this hypothesis is based on a previous study suggesting that visually-guided, robotic ankle training in a seated position can effectively change passive ankle stiffness.4 Hypothesis 3: The proposed robotic training will significantly improve paretic ankle motor control, postural balance control and locomotion in stroke survivors. The rationale for this hypothesis is based on our recent preliminary study demonstrating that robotic balance training that targeted altered 2D ankle stiffness in chronic stroke survivors improved lower extremity function, evidenced by less variable, smoother and faster ankle movement and improved clinical assessment scores for postural balance and walking.5

2. Background Information and Literature Review Health Benefits to Arizonans: Stroke is one of the leading and most lasting causes of long-term disabilities in Arizona. Notably, the incidence of stroke doubles in each successive decade after the age of 55.6 With the unprecedented growth of the older population in Arizona, effective and efficient stroke rehabilitation has become an important health, social and economic issue.7 We propose a paradigm-shifting robotic approach to advance current practices in balance training for stroke rehabilitation. The proposed approach would not only improve the overall lower extremity function of stroke survivors, but also improve the quality of their daily activities and reduce relevant healthcare costs. These benefits are well aligned with the Arizona Biomedical Research Centre’s mission to support innovative biomedical research to improve the health of all Arizonans.

Significance: Among the possible sensorimotor consequences of a stroke, impaired control of postural balance is a major cause of chronic disability that heavily impacts the quality of life of stroke survivors.8-10 This impairment makes daily activities more challenging and fatiguing and increases the risk of falls and fall-related injuries.11-14 While each joint in the lower extremity contributes to the neuromuscular control of postural balance, mounting evidence demonstrates that the ankle joint contributes to the maintenance of anteroposterior and mediolateral stability and is the first line of defense to external disturbances (i.e., ankle strategy).15-20 Importantly, most stroke survivors suffer from ankle impairments (e.g., contractures,21-23 spasticity24-27 and muscle weakness28-30) that contribute to the postural balance disorder. Hence, to advance stroke rehabilitation aimed at improving the control of postural balance, there is a significant need to understand how stroke alters ankle mechanics and to integrate this knowledge into the design and execution of training exercises to improve the ankle strategy in postural balance control.

PI Hyunglae Lee, Detailed Project Description, BPM000338, 6

RFGA NO. BPM000338

Current Challenges: Current clinical practices in stroke rehabilitation that have focused on improving the ankle strategy rely on subjective clinical assessment of ankle mechanics (e.g., Range of Motion,31 Modified Ashworth Scale,32 Manual Muscle Testing33), which often limits the effectiveness of rehabilitation. To overcome limitations of subjective clinical assessment, several studies have utilized robotic technologies and characterized altered ankle mechanics following a stroke.26,34-39 However, these characterizations were limited to a seated position and a single plane of motion (sagittal plane), thus preventing direct application to balance training exercises for stroke rehabilitation. Mainly due to the lack of appropriate research methodology, little is known about how multi-dimensional ankle mechanics are altered following stroke during postural balance.

On the other hand, several recent studies have demonstrated that joint-specific robotic rehabilitation for the paretic ankle effectively improves both postural balance and walking functions in stroke survivors.4,40,41 However, trainings in these studies were limited to isolated ankle motor control in a seated position without knowing altered ankle mechanics and thus might limit the degree of functional improvement. Given the importance of task-specific training in stroke rehabilitation,42-44 we expect that robotic ankle training will be more effective if performed during functional tasks such as upright postural balance and walking.

Solution and Impacts: This proposed research addresses these unmet needs by introducing an innovative robotic approach to identify altered ankle mechanics, particularly 2D ankle stiffness (elasticity) and damping (viscosity), during upright postural balance and to integrate this advanced knowledge into the design and execution of a patient-adaptive robotic balance training program.

This project will advance scientific knowledge on how patient-adaptive robotic balance training based on the quantification of altered ankle mechanics can improve the effectiveness of lower-extremity stroke rehabilitation.

Successful completion of this project will provide a new robotic approach to overcome limitations of current clinical practices that often lead to motor recovery plateau. The proposed robotic training can be also extended to benefit individuals with other motor impairments (e.g., Parkinson’s disease and multiple sclerosis) as well as elderly people at high risk of falls. Importantly, the proposed approach utilizing a patent-pending robotic platform will be filed for a new patent, which will impact the local economy by adding a new biomedical patent.

3. Preliminary Work This proposal stems directly from the PI’s previous research on the characterization of ankle mechanics using a wearable ankle robot45-51 and a robotic balance platform,52-56 which has made major scientific achievements related to the current proposal. Preliminary studies with stroke survivors have also confirmed that 2D ankle mechanics can significantly deviate from the unimpaired baseline, emphasizing the significance of characterizing multi-dimensional viscoelastic ankle mechanics in stroke survivors.3,57 More recently, the PI has performed two pilot studies in collaboration with Drs. Kinney and Eikenberry, demonstrating that: 1) ankle stiffness in chronic stroke survivors was significantly different from the unimpaired baseline and 2) robotic balance training targeting altered 2D ankle stiffness improved the overall lower extremity function in chronic stroke survivors, which is an important motivation for the current proposal.5 PI and Team Qualification: The PI has gained broad research experience in robotics, biomechanics, system identification and neurorehabilitation at three institutions for the past 10 years, uniquely qualifying him for the proposed research. He has made critical contributions to the field, including human-interactive robots and controllers, system identification methods for the human neuromuscular system, and the quantitative characterization of human joint/limb mechanics and stability. Moreover, in addition to his prolific publication of research outcomes in highly selective journals (e.g., IEEE Transaction on Neural Systems and Rehabilitation Engineering),45-52,58-62 he has recently received a National Science Foundation Faculty Early Career Development Program (NSF CAREER) award from the program of Disability and Rehabilitation Engineering.

This project also involves uniquely qualified medical practitioners. Dr. Kinney (consultant) is a physician specializing in stroke rehabilitation at Mayo Clinic Phoenix, AZ. She has 17 years of experience in the clinical assessment and treatment of postural balance and gait in patients recovering from stroke. Dr. Eikenberry (co- I) is a physical therapist specializing in stroke rehabilitation at Midwestern University in Glendale, AZ. She has 12 years of experience in stroke assessment and rehabilitation and has worked within Midwestern University, Mayo Clinic and within the national American Physical Therapy Association to improve the objective nature of functional balance measures routinely utilized in stroke rehabilitation. The PI has been actively collaborating with Drs. Kinney and Eikenberry for the past 2 years to design studies foundational to the current proposal, to run pilot studies to demonstrate their feasibility, and to present and publish preliminary results.3,5

PI Hyunglae Lee, Detailed Project Description, BPM000338, 7

RFGA NO. BPM000338

4. Experimental Methodology Participants and Inclusion Criteria: From stroke databases at the Mayo Clinic, Midwestern University and physical therapy centers located in the Phoenix area, we will select 20 chronic stroke survivors. Only those with a left hemisphere lesion at least 6 months prior to participation will be considered. In addition, we will use the Berg Balance Scale (BBS) score,63,64 Modified Ashworth Scale (MAS)32 and comfortable gait speed65,66 to select subjects with mild or moderate impairments in postural balance (BBS ≥ 41), spasticity at the ankle (1 ≤ MAS ≤ 2) and limited community ambulators (comfortable gait speed ≥ 0.4 m/s). Other inclusion criteria are as follows: 1) absence of visual impairment, peripheral vestibular disorders and/or loss of ankle proprioception;

2) an ankle range of motion greater than 5° in dorsiflexion, plantarflexion, inversion and eversion; 3) absence of significant motor impairment in the unimpaired limb; 4) absence of any acute or chronic painful condition; 5) absence of severe concurrent medical problems (e.g., cardiorespiratory impairment); and 6) ability to stand unassisted for 5 minutes. Twenty age-matched, unimpaired controls with no history of neuromusculoskeletal disorders in the lower extremity and back will be recruited from local recreational centers (e.g., YMCA).

Experimental Setup: As a centerpiece of the proposed research, we will use a patent-pending, dual-axis robotic balance platform,52-54 which covers a wide range of ankle motion in 2 degrees-of-freedom (DOFs), specifically, 30° in the dorsiflexion-plantarflexion (DP) in the sagittal plane and 20° inversion-eversion (IE) in the frontal plane (Fig. 1A). The platform is capable of providing precisely-controlled position perturbations, measuring the corresponding ankle torques, and simulating a wide range of stiffness- and damping-defined mechanical environments in 2 DOF of the ankle. We have successfully used this platform to quantify 2D viscoelastic ankle mechanics during upright postural balance in unimpaired individuals under various mechanical environments, from compliant to rigid environments.52,55,56,67 A 3D motion capture system (8 cameras) will be used to record lower body kinematics. We will analyze kinematic data not only of the ankle to quantify 2D viscoelastic ankle mechanics but also of the knee and hip to investigate any confounding effects of the lower extremity kinetic chain on the quantification of ankle mechanics. A visual feedback display will be used to control upright posture during the quantification of ankle mechanics and to visually guide ankle movement during balance training. Further, a weight scale will be used to control the equal weight distribution between both legs. In addition, to monitor muscle fatigue throughout the experiment, surface electromyography (EMG) sensors will be attached to the belly of four major ankle muscles: tibialis anterior, soleus, medial gastrocnemius and peroneus longus.68 To ensure safety, each subject will wear a safety harness fastened to a body weight support system. The experiment will be performed at ASU, and the required IRB will be approved prior to the start of the study.

Research Design and Methods for Aim 1: The objective of this aim is to accurately quantify 2D viscoelastic ankle mechanics in 20 chronic stroke and 20 age-matched, unimpaired control subjects. We will quantify ankle stiffness and damping in both the sagittal and frontal planes during upright postural balance. Based on our preliminary results, we hypothesize that both 2D ankle stiffness and damping are significantly altered following stroke.3 Outcomes of this task will be utilized in the subsequent tasks to design (Aim 2) and execute robotic balance training (Aim 3).

In the experiment, subjects will stand upright with the right foot (affected side) on the robotic platform and the left foot on the weight scale beside the platform (Fig. 1B). Subjects will be instructed to maintain the reference 2D center-of-pressure (COP) in a space consisting of IE and DP axes, measured during quiet upright standing at the beginning of the experiment. The visual feedback display will inform subjects in real-time of their 2D COP in respect to the reference. Once the subject maintains the reference COP for a randomized time interval of 0.5−2 s, mild random position perturbations (pseudo-random binary sequence with an amplitude of 3°, a speed of 100°/s and a switch rate of 100 ms) will be applied. Perturbations will last 5 s, and a total of 20 trials will be completed in each direction of ankle movement (i.e., DP and IE), resulting in 40 total trials. The order of perturbation directions will be randomized. Experiments will be conducted in blocks of 20 trials, with a minimum rest period of 3 mins between blocks. Each trial will be completed in ~10 s, requiring a total experimental time of ~10 mins.

Fig. 1: (A) Dual-axis robotic balance platform.

(B) Experimental setup.

PI Hyunglae Lee, Detailed Project Description, BPM000338, 8

RFGA NO. BPM000338

We will quantify 2D viscoelastic ankle mechanics (i.e., ankle stiffness and damping in DP and IE) using an advanced system identification method that integrates a short segment algorithm69,70 with correlation-based methods from the PI’s previous studies.47-49 This integrated algorithm allows for reliable and robust identification, even with a limited number of data and in the presence of a poor signal-to-noise ratio. As such, this integrated algorithm outperforms classical time-invariant71 and ensemble-based time-varying algorithms72 and allows us to simultaneously quantify 2D ankle stiffness and damping using small amounts of data.

Research Design and Methods for Aim 2: The objective of this aim is to develop a new robotic balance training program based on quantification results of altered ankle mechanics in stroke survivors compared to the age-matched, unimpaired controls. Building upon our preliminary study demonstrating that a 5-week robotic balance training program targeting altered ankle stiffness significantly improved the overall lower extremity function in chronic stroke survivors,5 we will further develop this training program by targeting both altered 2D ankle stiffness and damping and integrating assist-as-needed control and patient-adaptive control schemes (Fig. 2). The proposed method will be implemented on the dual-axis robotic balance platform so that training is conducted in a functional task situation (i.e., upright postural balance).

The robotic balance training consists of visually-guided target reaching tasks on the robotic platform simulating stiffness- and damping-controlled mechanical environments. The stiffness and damping values of the robotic platform (kr and br) will be initially determined as those of age-matched unimpaired controls (ko and bo) but will be adjusted in a patient-adaptive manner, depending on the assessment of the patient’s motor performance throughout the training exercise. We chose visually-guided target reaching ankle movement as a training strategy because it is simple enough to be conducted by stroke survivors, yet its efficacy in the rehabilitation of the lower extremity, including the ankle, has been historically demonstrated.4,40,73 The patient will perform target reaching from the neutral ankle position (measured during upright standing on a rigid surface) to one of four different targets in dorsiflexion, plantarflexion, inversion and eversion directions (Fig. 3).

Each targeted ankle movement will be completed according to the assist-as-needed control scheme to promote patient engagement.74 The target movement trajectory will be prescribed by a minimum jerk trajectory (θm,j).75 Assistive torque (τassistive) will be provided to the ankle only when the current ankle position (θ) is behind the current desired ankle position (θm,j). When the current ankle position is between the desired position (θm,j) and the target position (θtarget), no assistance will be provided, and patients will only experience the torque from the robotic environment (τmech), defined by kr and br. If the current position overshoots the target position, resistive torque (τresistive) will be provided. In summary, active torque (τactive) that patients will experience is as follows:

(1)

(2)

(3) where td is the desired duration of the movement to the target. The optimal values for controller stiffness kassistive and kresistive will be determined throughout the funding period of this proposal.

Further, the adaptive impedance control scheme will be integrated to adjust the amount of robotic assistance in a patient-adaptive manner. A performance observer will monitor the amount of robotic assistance during targeted ankle movements and adjust impedance of the robotic platform (zr; the stiffness [kr] and damping [br]) τ active = τ assistive +τmech (θ <θm, j ) τmech (θm, j ≤θ ≤θt arget ) τ resistive +τmech (θ >θt arget ) τ assistive = −kassitive (θ −θm, j ) τmech = −krθ −br !θ τ resistive = −kresistive (θ −θt arget ) θm, j =θt arget 10( t td )3 −15( t td )4 +6( t td )5

Fig. 3: Visual feedback for targeted ankle movements. Final targets are located at 5° from the neutral position. An example in the dorsiflexion direction is shown.

Fig. 2: Summary diagram of the patient-adaptive robotic balance training based on the assist-as-needed control scheme.

PI Hyunglae Lee, Detailed Project Description, BPM000338, 9 once each training session is completed. If substantial robotic assistance is observed in the current session, the mechanical environment in the next session will be adjusted to be more compliant (lower zr). This adjustment is expected to help patients better complete targeted ankle movements. On the other hand, if little robotic assistance is observed in the current session, the mechanical environment in the next session will be adjusted to be more challenging with increased impedance of the robotic platform (higher zr). An adaptation control law to determine the amount of adjustment (Δz ; Δk and Δb) is defined as follows:

(4)

1, respectively. P is initially set to 0.2 but can be changed throughout the funding period of this proposal.

Research Design and Methods for Aim 3: The objective of this aim is to evaluate the effectiveness of the proposed 5-week, patient-adaptive, robotic balance training program for improving overall lower extremity function in chronic stroke survivors. Based on our preliminary results, we hypothesize that the robotic balance training program will not only change the altered viscoelastic ankle mechanics in stroke survivors toward the unimpaired baseline but also improve paretic ankle motor control, postural balance control and locomotion.

Each stroke subject will participate in the 5-week robotic balance training program consisting of 10 sessions (2 sessions/week; 1 hr/session) (Fig. 4). Each training session will be conducted following the approach described in Aim 2. To evaluate the effectiveness of the proposed training, we will use both quantitative outcome measures and popular clinical measures at three different time points: pre-training (baseline), post-training (5-week) and follow-up (3-month). We will quantify the variability, smoothness and speed of paretic ankle movement throughout the training sessions and investigate any changes in these measures. We will also evaluate functional improvement in postural balance and walking by assessing BBS63,64 and 10 Meter Walk Test (10 MWT)76 scores, respectively. In addition, to evaluate the effectiveness of the proposed training in correcting altered ankle mechanics toward the unimpaired baseline, we will quantify 2D ankle stiffness and damping at the same 3 time points.

Each training session will be composed of 8 blocks of movement trials. A total of 40 trials will be completed in each of the 4 ankle movement directions (i.e., dorsiflexion, plantarflexion, inversion and eversion), resulting in a total of 160 trials. During each trial, subjects will stand upright on the robotic platform (Fig. 1B) and perform visually-guided target reaching tasks by controlling ankle position to the randomly selected targets around the neutral ankle position (amplitude = 5°). Visual feedback will show their current and desired ankle positions at each time step as well as the neutral and final target positions (Fig. 3). Each block consists of 20 trials, with 10 in each of the 2 movement directions for the selected plane. For DP movement trials in the sagittal plane, IE movement in the frontal plane will be restricted and vice versa. The blocks will be alternated between the two planes. Each trial is expected to be completed in ~15 s, and a minimum rest period of 3 mins between blocks will be provided to prevent fatigue, requiring a total experimental time of ~1 hr. Once each training session is completed, the stiffness and damping of the robotic platform will be adjusted according to the adaptation control law described in Aim 2 (Eq. [4]). The control law will apply to each ankle movement direction separately.

Innovations: A robotic approach integrating the advanced system identification method into the patent-pending, dual-axis robotic platform will provide us an innovative solution to quantify 2D viscoelastic ankle mechanics (ankle stiffness and damping) during upright postural balance, not achievable with current state-of-the-art technologies. Application of this advanced knowledge to postural balance training will uniquely determine mechanical environments of the robotic platform (i.e., stiffness and damping values of the robotic platform) that stroke survivors will interact with during balance training. Furthermore, integration of assist-as-needed control and adaptive control schemes with the proposed training exercise will not only promote patient engagement but also adaptively adjust the amount of robotic assistance in a patient-specific manner throughout the exercise, which is not currently achievable in balance training practices for stroke rehabilitation.

Fig. 4: Summary diagram of the 5-week balance training program.

PI Hyunglae Lee, Detailed Project Description, BPM000338, 10

RFGA NO. BPM000338

5. Data Analysis Data Collection: We will use two data acquisition boards: one for the robotic platform and the other for human participants. The first board will collect kinematic and force data at the robotic platform at 2 kHz. The second board will collect 3D lower body kinematics at 200 Hz and weight distribution between the legs and EMG data at 2 kHz. All data will be synchronized using a shared triggering signal. Kinematic and force data will be low-pass filtered using a bi-directional 4th order Butterworth filter with a cut-off frequency of 20 Hz. We will quantify the degree of muscle fatigue by calculating the median frequency shift of the EMG power spectrum.68 Statistical Analysis and Expected Outcomes: The Hypothesis 1 is that 2D viscoelastic ankle mechanics in stroke survivors are significantly different from those in age-matched controls. We will evaluate this hypothesis for each stroke subject individually since altered ankle mechanics are expected to be highly variable across subjects. We will use Welch’s unequal variance t-test and perform separate analyses for each ankle parameter (stiffness and damping) and each movement direction (DP and IE). We expect that each stroke subject will have significantly altered (p<0.05) ankle stiffness and damping compared to those in controls. The Hypotheses 2 and 3 are that the proposed robotic training not only changes altered viscoelastic ankle mechanics toward the unimpaired baseline but also improves overall lower extremity function in stroke survivors. We will assess these hypotheses using a one-way repeated measures ANOVA with an independent variable of testing time points (pre-training, post-training and 3-month follow-up). Four dependent variables for ankle mechanics (ankle stiffness and damping in DP and IE) and five dependent variables for ankle motor control (variability, smoothness and speed), postural balance control (BBS score) and locomotion (10 MWT score) will be analyzed. Post hoc analyses will be completed using Tukey’s honestly significant different test. At the conclusion of the 5-week robotic training, we expect to observe significant changes in 2D viscoelastic ankle mechanics toward the unimpaired baseline. We also expect significant improvement in both quantitative measures (variability, smoothness and speed in paretic ankle motor control) and clinical measures (BBS and 10 MWT). Further, these changes in ankle mechanics and functional outcomes are expected to be relatively well maintained over the post-training period, resulting in a statistical significance at the 3-month follow-up in reference to the pre-training baseline. The level of statistical significance for all measures will be set at p<0.05.

Potential Difficulties and Solutions: Since our preliminary studies have demonstrated the reliability of the robotic quantification approach and the feasibility of the robotic balance training, we do not foresee any major difficulties.3,5 However, we might encounter a challenge in cases where stroke participants have severely

adaptive control algorithm (Eq. [4]) will address this issue over the training, it would be more efficient if the training started with a less challenging task. To this end, mechanical environments could be redefined with a weighting factor (w) as wzr, where 0< w ≤1. The more severe the alteration in ankle mechanics, the smaller the value w. A criterion to select w will be determined throughout the funding period of this proposal.

Safety Features: First, subjects will wear a safety harness. Second, the robotic platform has multi-level safety features. It is equipped with two emergency stop buttons, which are accessible by both researchers and participants. The buttons override the system and immediately stop all motor force once either is pressed. The system is also programed with safety thresholds for velocity, position and the torques generated by the motors.

6. Timeline

7. Evaluation We will evaluate the reliability of our ankle mechanics quantification (Objective 1) by calculating the variance accounted for between the measured and predicted ankle torques calculated by the convolution of the measured ankle kinematics and identified ankle parameters (i.e., stiffness, damping and inertia). We will evaluate the effectiveness (Objectives 2 & 3) of the proposed robotic balance training by rigorously testing the three hypotheses with the statistical analysis described in Section 5.

PI Hyunglae Lee, Detailed Project Description, BPM000338, 11

RFGA NO. BPM000338

DETAILED PROJECT DESCRIPTION (PROJECT RESOURCES)

1. Laboratory The research will be performed at the Neuromuscular Control and Human Robotics Laboratory, directed by Dr. Lee (PI). The lab is approximately 800 sq. ft. and located at Arizona State University (ASU) ECE 116, adjacent to Dr. Lee’s office (ERC 373). This space contains computers and software that can be utilized for mechanical design, data processing, analysis, and word processing.

Dr. Lee is also managing a motion analysis laboratory for biomechanics and human movement science studies, located in the ECC 117 at ASU. All experiments will be conducted in this space, which contains all of the equipment, computers and software for the proposed research. We are well equipped with excellent laboratory space, facilities and equipment to successfully conduct the proposed research.

2. Clinical Dr. Lee has close relationships with clinicians at Mayo Clinic, Arizona (Carolyn Kinney, MD, Dean Wingerchuk, MD, Ibrahim Aksoy, MD PhD, and Megan Eikenberry (also has a position at Midwestern University), PT DPT NCS), located within 25 miles from ASU. This hospital works with both in and outpatient stroke survivors.

Especially, Dr. Lee has been actively collaborating with Drs. Kinney and Eikenberry for the two past years to run pilot studies and 1) demonstrated the effectiveness of the dual-axis robotic platform for the quantification of multi-dimensional ankle mechanics in stroke and multiple sclerosis patients, and 2) confirmed the feasibility of a robotic balance training approach that became a foundation of the proposed research in this proposal.

Drs. Megan Eikenberry and Carolyn Kinney will participate in the proposed research as a Co-I and a consultant, respectively. In the proposed study, Dr. Eikenberry will recruit stroke subjects, complete clinical assessments, and assist with human experiments at ASU. Dr. Kinney will assist in the design of experimental protocols, as needed, interpretation of the results with emphasis on clinical significance and translation to clinical application, and manuscript preparation.

3. Animal Not applicable.

4. Computer The Neuromuscular Control and Human Robotics Lab has several desktop PCs for data acquisition and analysis, which include multipurpose data acquisition cards (National Instruments, TX) and all necessary software (e.g. Microsoft Office, MATLAB, SPSS). Three computers are also equipped with 3D CAD software (Solidworks) for mechanical design. All computers are fully networked with high-speed Internet connections.

Skype and Zoom teleconference software are on hand for all members of the research team at ASU, Midwestern University, and Mayo Clinic. This computing environment, which enables efficient data handling, processing, discussion, and reporting, has been extensively used for the past years among the team members (Drs. Lee, Eikenberry and Kinney).

5. Office Dr. Lee has a 150 sq ft office in ERC 373, adjacent to his labs. Student office space is provided within the main research laboratory (ECE 116A). The motion analysis lab (EEC 117) also has an office space for students. All offices are well equipped with desks, chairs, filing cabinets, PCs, and hardwired high-speed Internet, wireless Internet. This office environment enables efficient communication among members of the research team.

6. Major Equipment Dual-Axis Robotic Balance Platform: We have developed and validated a robotic platform capable of actuating 2 DOF of the ankle (dorsiflexion-plantarflexion (DP) in the sagittal plane and inversion-eversion (IE) in the frontal plane) and simulating various mechanical environments, from compliant to rigid environments.

Two servo motors (AKM 43L, 33H; Kollmorgen, VA) and drivers (AKD-01206 and AKD-00606; Kollmorgen, VA) are used to actuate the robotic system. It allows for a wide range of motion of 20° from horizontal in the sagittal plane and 10° in the frontal plane. It can provide rapid ramp perturbations of the average speed up to 100°/s and the maximum speed of 200°/s in both directions and peak torque of 400 Nm and 200 Nm in DP and IE directions, respectively. The platform is designed to be robust even in a substantial loading condition

PI Hyunglae Lee, Project Resources, BPM000338, 12

RFGA NO. BPM000338

when a 100 kg subject acts on the edges of the platform during walking. A dual-axis force plate (9260AA6;

Kistler, Switzerland) is mounted on the top plate to record the ground reaction forces and calculate torques at the ankle in 2 DOFs. Both angles and torques will be recorded at 2 kHz. The error of position control is less than 0.1° in both DOFs. A paper on the development and validation of this robotic platform has been published at IEEE/ASME Transaction on Mechatronics (2019). This equipment is currently patent-pending.

3D Motion Capture System: The Flex 13 motion capture system (OptiTrack, OR) is a set of cameras and associated software designed to track and measure the location of small, passive, retro-reflective spherical markers that are placed at various anatomical landmarks on the body of human subjects performing various movement tasks such as walking, running, balance, and other dynamic activities. The system consists of 8 cameras (1.3 Mega Pixels, expansive 56° field of view, 120 fps) to cover the entire lower body from the hip to the ankle.

EMG Data Collection System: The EMG system Trigno Wireless 16 (Delsys Inc., MA) is a revolutionary hybrid EMG/Movement sensor with motion artifact suppression. It includes 16 surface wireless electrodes that have embedded signal conditioning and amplification, as well as 3D accelerometers for removing motion artifact in EMG signals. Moreover, the acceleration measurements are included in the data acquisition in order to be easily incorporated with motion capture systems. It has 16 EMG channels and 48 accelerometer channels, selectable ± 1.5 g and ± 6 g range of acceleration. Parallel-bar technology for high-fidelity EMG signals. 16-bit resolution, 2 kHz sampling rate. EMG data will be used in the proposed research to monitor muscle fatigue.

Visual Feedback Display: During quantification of 2D ankle mechanics (Aim 1), a visual feedback will enable human subjects to maintain the same upright posture throughout the experiments. Subjects will receive real-time data regarding the current center-of-pressure (COP) in a 2D space consisting of DP and IE axes as well as the reference COP measured during quiet standing at the beginning of the experiment. During balance training (Aim 3), the visual feedback will allow for visually-guided target reaching ankle movement. The feedback display will show current and desired ankle positions at each time step as well as the neutral and final target positions. An LCD monitor (22") will be used as a visual feedback display and properly installed to allow subjects accurately see the provided information.

Safety Harness System: To ensure safety of subjects throughout the experiments, each subject will wear a safety harness fastened to the body weight support system (LiteGait, AZ). This system can support a subject up to 6'7"−6'11" tall and 500 pounds.

Data Acquisition and Processing Software: ASU supplies the latest Mathworks tools (Mathworks Inc, Natick, MA) including MATLAB, Simulink, Data Acquisition Tool box, Real Time Workshop and xPC Target for data acquisition and processing.

Major equipment described here is installed in the motion analysis lab (ECC 117) and collectively constitutes an environment that has the full range of capabilities needed to successfully complete the proposed research.

All this equipment is exclusively assigned to the PI and will be always available for the proposed research.

7. Other On-campus Resources: Dr. Lee has direct access to the Fulton Schools of Engineering machine shop, located in PSYN 118 (<5 minutes walking distance from the PI’s laboratories and office) and manned with machinists trained in CAD manufacture. He also has access to the Fulton Schools of Engineering 3D Print and Laser Cutter Lab, located in ECF 103 right next to the PI’s laboratories. This facility is equipped with high-end 3D printers and laser cutter devices and will be available anytime during weekdays throughout the year.

Institutional Commitment to Early Stage Investigator: Dr. Lee has received extensive support from his department (Mechanical Engineering in the School for Engineering of Matter, Transport, and Energy (SEMTE)) including but not limited to: tenure-track position, institutional equipment start-up funds, administrative support (pre-award, post-award, IT, business,…