Citation
Acute biomechanical effects of a therapeutic exoskeleton during over-ground walking in healthy controls and individuals post-stroke

Material Information

Title:
Acute biomechanical effects of a therapeutic exoskeleton during over-ground walking in healthy controls and individuals post-stroke
Creator:
Beitter II, John M.
Publication Date:
Language:
English

Notes

Abstract:
This study evaluated the effects of a single-limb, therapeutic exoskeleton on the biomechanics of over-ground walking in groups of healthy controls and individuals post-stroke. We investigated knee biomechanics to inform future designs of this, or other, exoskeletons. Participants were assessed clinically, which included the Berg Balance Scale and a 6-minute walk test. They were given the opportunity to familiarize themselves with the device before retesting while wearing the device and participating in biomechanical testing to determine the effects of various device configurations including: no actuation (free-swing), assistance, and threshold percentage. We discovered that there was an effect of wearing the device, primarily in the kinematics of the knee during over-ground walking, and the effect was differentiated by group. The effect is characterized by increased flexion during stance in the group post-stroke and decreased flexion during pre-swing and swing in the control group. There were no significant group x condition, group x assistance percentage, or group x threshold percentage interactions in the internal knee moment, however at initial contact and during pre-swing there were significant threshold percentage and condition main effects, respectively. The effects of changing the assistance and percentage threshold were discussed to optimize design characteristics and enhance clinical utility. ( en )
General Note:
Awarded Bachelor of Science in Biomedical Engineering, magna cum laude, on May 8, 2018. Major: Biomedical Engineering
General Note:
College or School: College of Engineering
General Note:
Advisor: Daniel Ferris. Advisor Department or School: BME

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright John M. Beitter II. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

UFDC Membership

Aggregations:
UF Undergraduate Honors Theses

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

Acute biomechanical effects of a therapeutic exoskeleton during over ground walking in hea l thy controls and individuals post stroke Author: John Beitter Oral Defense: April 18, 2018 Supervisory Committee: Dr. Daniel Ferris, Dr. Jennifer Nichols, Dr. Carolynn Patten

PAGE 2

Abstract This study evaluate d the effects of a single limb, therapeutic exoskeleton on the biomechanics of over ground walking in groups of hea l thy contr o ls and individuals post stroke. We investigated knee biomechanics to inform future designs of this or other exoskeletons Participants were assessed clinically, which included the Berg Balance Scale and a 6 minute walk test T hey were given the opportunity to familiarize themselves with the device before retesting while wearing the device and participating in biomechanical testing to determine the effects of various device configurations including: no actuation (free swing), assistance, and threshold percentage We discovered that there was an effect of wearing the device, primarily in the kinematics of the knee during over ground walking and the effect was differentiated by group The effect is characterized by increased flexion during stance in the group post stroke and decreased flexion during pre swing and swing in the control group. There were no significant group x condition, group x assistance percentage, or group x threshold percentage interactions in the internal knee moment, however at initial contact and during pre swing there were significant threshold percentage and condition main effects, respectively. The effects of changing the assistance and percentage threshold were discussed to optimize design characteristics and enhance clinical utility.

PAGE 3

Introduction Stroke typically results in impaired motor function, in particular the impairment of an individual s walking ability [1] Statistics as of 2018 state that each year 610,000 new people experience a stroke, and 185,000 people fall victim to a recurrent attack in the United States [2] Despite up to 85% of the se victims regain ing their ambulatory ability after six months c haracteristics of their gait pattern include reduced knee flexion in swing and reduced internal knee moments during stance [1] Reduced swing phase knee flexio n is thought to be a result of mult iple mechanisms: weakness i n hip flexors during swing hyperactivity of the quadriceps during swing and reduced plantar flexor act ivity during mid stance [3], [4] Robotics have been used to augment normative human ability and counteract the negative effects of stroke in order to improve mobility [5] Two categories of human augmentation devices are being investigated as a potential solution to gait impairment post stroke : assistive robotic exoskeletons, robotics which supplement existing ability and are intended for use in daily life and therapeutic exoskeletons robotics which are used to supplement or replace traditional therapy [5] This paper evaluates one exoskeleton the Tibion PK100 bionic leg (TBL ), which was designed as a therapeutic device for the lower extremity [6] The TBL is actuated at the knee providing resistance to knee flexion or assistance to knee extension during the stance phase of the gait cycle while walking down stairs, or while r ising from a seated position ( For mor e information about the Tibion bionic l eg see [7] ). T he goal of this study is to evaluate the acute effec ts of the TBL t pattern, focusing on the knee where the device is actuated Engineers have commonly attempted to evaluate the effectiveness of their device s by performance measures such as portability, durability and operability [7] Portability durability, and operability of the TBL were evaluated by Horst [7] This evaluation also includes ability of the orthosis to i is rather qualitative, and for therapeutic exoskeletons effectiveness can be difficult to observe due to the fact that the device is not used in daily life Instead, it is necessary to evaluate the perfor mance of the device by investigating the effects of the device, both in an acute setting and after treatment Effectiveness of the TBL has already be e n evaluated in various case studies by comparing human performance after therapy [8 ] [11] Case studies found that the device

PAGE 4

improved the users balance after therapy as measured by the Berg Balance Scale (BBS) [8], [9] A l arger study confirmed the outcome that BBS measurements favored a TBL supplemented therapy group post stroke [10] O ne study did not measure balance with the BBS and did not observe an im proved balance but the researchers acknowledged that the tests utilized might not have been sensitive enough to [11] They also recognized that the strength of the knee extensors was not assessed post training but it was clear the device promotes functional improvements [11] The most recent study, performed by Li et al. in 2015 on three individuals post stroke, measure the internal knee moment pre and post training and found a significant increase in the mean knee extension moment during the stance phase of the gait cycle [9] Therefore, research exists about the outcomes of treatment supplemented by the device, but a need still exists to quantify the performance of the TBL acutely during walking. What immediate eff ect does wearing the device have on the biome chanics of the knee during over ground walking ? How are the effects of the device changed at various assist ance and threshold percentages? Finally, the de vice was primarily intended to a ffect the biomechanics of the knee during stance, but how do these effects in stance and the added loa d to the leg a ffect the biomechanics of the knee during swing? Evaluation of the acute effects of the exoskeleton and t herapeutic outcomes could inform future designs of ro botics specifically exoskeletons, for stroke rehabilitation benefitting engineers, clinicians, and patients. Hypothesis Given the assistance to knee extension provided by the device and the added load to the device wearing leg f or the control group we hypothesized that wearing the TBL would result in decreased peak internal knee extensor moment during stance and decreased knee flexion during swing. However, c onsidering the findings of Li et al 2015, w e hypothesize d that wearing t he TBL would result in increased peak internal knee extensor moment during loading response and mid stance and decrease d knee flexion during swing for the group post stroke [9]

PAGE 5

Methods Participant Information Twenty nine indiv iduals participated ( mean age : 55.4 SD 13.9 years ) Nineteen of the individuals had experienced at least one, but no more than three, unilateral strokes which produced either subcortical or cortical lesions resulting in hemiparesis. These individuals had at least a minimal ability to ambul ate independently with or without an assistive device over level ground. Five i ndividuals could not safely participate without their assistive device, so they were excluded from the kinetic analysis. Kinetic data from the device leg of t wo controls were excluded from the analysis due to invalid force plate strikes. R elevant group characteristics are reported in Table 1 below including clinical assessment s described in the protocol Table 1. Participant Demographics Control Stroke Participants (n) 10 19 Age (yrs) 53.1 SD 17.1 57.1 SD 12.3 Sex (M/F) 5 / 5 14 / 5 Height (cm) Mean 172.2 SD 13.0 175.1 SD 7.4 Range 157.5 189.5 164 192 Weight (kg ) Mean 79.0 SD 20.3 88.5 SD 14.0 Range 90 226.4 135.5 243.2 Chronicity (mo) 73.1 SD 41.4 Mini BEST (/32) 17.8 SD 7.2 Berg Balance Scale (/56) 46.2 SD 7.60 6MWT (m) 127.6 SD 132 LE Fugl Meyer Motor Assessment (/34) 22.7 SD 6.2

PAGE 6

Robotic Exoskeleton The TBL 1 is a n electrically actuated therapeutic exoskeleton created to assist in the rehabilitation of persons with impaired lower extremity function. Five parameters are input by the clinician the weight of the user, the threshold or minimum percentage of total body weight force required for the user to produce and be sensed by the foot sensor before the device provides assistance, the assistance factor or the amount of assistance as a percentage of the maximum output provided by the device during knee extension, the resistance factor or the amount of resistance as a percentage of the weight of the user provided by the device during knee flexion, and the extension range of motion limit A foot switch in the shoe insert detects if the foot is in contact with the ground. Sensors i n a shoe insert detect if the threshold has been met, and an onboard microprocessor initiates device actuation. The total mass of the device insert, and battery is 3.5 kilograms. Protocol The study protocol consisted of 4 sections including: (1) c linical assessments, (2) d evice f amiliarization (3) assessment of spatio temporal gait parameters and (4) b iomechanical testing. Clinical assessment s for the group post stroke included the : Late Life Function & Disability Instrument (LLFDI) [12] [13] Berg Balance Scale (BBS) [14], [15] Lower Ex tremity Fugl Meyer Assessment (LE FMA) [16] Modified Ashworth Scale (MAS) [17] [18] mini Balance Evaluation Systems Test (mini BEST) [19] [20] [21] and 6 minute walk test (6MWT) [22] Following clinical assessments, both groups of participants entered a familiarization period which co nsiste d of fitting the device, weight bearing functional mobility activities, and optimization of the five device inputs by the clinician. Weight bearing functional mobility activities included walking indoors on a level surface, outside on various surfaces, negotiating curbs and around other obstacles. Participants also perform ed sit to stand transfe rs from different seat heights and practice d walking up and down stairs. Finally the participants completed bal ance activities, such as single limb stance and assuming a tandem stance. At a maximum of 30 1 The Tibion PK100 Bionic Leg Orthosis, Version 0.5.5, Alter G Corporation, Fremont, CA.

PAGE 7

minutes, this period was used to instill confidence in the participants while walking with the device D uring this peri od the participants post stroke did not wear their ankle foot orthosis if possible Participants with hemiparesis wore t he device on the ir paretic leg, and the control group wore the device on a random ly select ed leg. A fter the familiarization period the MiniBEST and 6MWT clinical assessments were repeated while wearing the device. C ontrols and participants affected by stroke then walk ed on the GaitRite a computerized temporal and spatial gait analysis system while wearing the device at various predetermined device setti ngs (described below Table 2 ) to determine their self selected walking speed (SSWS) with the device. SSWS was determined using the average walking speed acro ss all configurations Both groups also walk ed on the GaitRite without the device to determine the ir SSWS without the device. Controls whose self selected walking speed was >1.00 m/sec were constrained to walk ing at 1.00m/sec during the biomechanical testing section. Each device configuration was tested twice Table 2: Device Settings at Configurations Tested During the biomechanical testing period participants walked over level ground without the device, and with the device configured in 6 different conditions In the first condition, to allow for evaluation of added mass and inertia of the device. The remaining trials were te sted in a random order Settings 1, 3, and 5 scaled the assistance percentage of the device linearly while holding the threshold constant Settings 7, 3, and 6 scaled the threshold percentage linearly while holding the assistance constant The resistance f actor was kept constant throughout the experiment at the default setting for the TBL, A minimum of three trials was collected for each condition. The number of steps processed varied per subject for kinematics ( mean 15 SD 6 ) and for kinetics (mean 4 SD 2) 1msWS0 (free swing) 1msWS1 1msWS3 1msWS5 1msWS6 1msWS7 Threshold 95 30 30 30 40 20 Assistance 10 30 50 70 50 50 Resistance Med Med Med Med Med Med

PAGE 8

Equipment Marker position data was sampled by 12 Vicon cameras at a frequency of 200 Hz (Vicon Motion Systems, Oxford, UK). Force plate measurements were sampled at 2000 Hz (ADAL Medical Developments, France). The marker set in cluded 11 clusters, 50 in dividual, and 80 total markers, with adjustments to marker positions made to account for the TBL. An illustration of the marker set, including descriptions of the placement of each marker and cluster is provided in Appendix A Clus ters of three markers were placed on the upper arms and forearms, and both feet. Clusters of four markers were used on both the thighs and shanks and a five marker cluster was placed on th e pelvi s During trials in which the participant was wearing the device, the marker set was adjusted by replacing the clusters of four markers on the thigh and shank with custom clusters of three markers. Finally, a spring loaded digitizing pointer was used during static trials to identify the medial and lateral knee and ankle ( Vicon Motion Systems, Oxford, UK ) [23] The TBL did not interfere with any markers or clusters a fter this adjustment was made. Clusters were secured in pla ce by elastic wraps, or Velcro. Data Processing and Analysis Participants were studied while walking over ground using instrumented 3D motion analysis Two separate biomechanical models were used to calculate kinematics and kinetics in Visual 3D ( V3D, C Motion, Gaithersburg, MD). Each participant had one model for conditions in which they wore the device and another model for conditions in which they did not wear the device. Joint centers were calc ulated using standard V3D modeling procedures [24] The hip joint center calculation used the Co da p elvis method (Charnwood Dynamics Ltd., Leicestershire, UK) [25] The knee joint center was calculated as the midpoint between the medial and lateral knee makers. The ankle joint center was calculated as the midpoint between the medial and lateral ankle markers. The Tibion outputs via Bluetooth, data for the torque that it applies to it s frame The events recorded by the foot switch of the TBL and the weight on the heel measured by the foot sensor were used to determine the time offset between ground reaction force measurements and Tibion torque outp ut with custom MATLAB (The Mathworks, Inc., Natick, MA) code. After

PAGE 9

adjusting the offset, the torque output was then time normalized to match the sampling frequency of the ground reaction force data This was imported to V3D, and t he torque measured by the device was applied to the model as a force couple These external forces resulted in a torque about the knee joint center which is equivalent to the measured output of the device. After V3D modeling and processing steps, ground reaction forces were filtered using 1 st order low pass B utterworth filter with a 10Hz cut off frequency in Matlab. Kinematics data were filtered using 1 st order low pass Butterworth filter with a 6 Hz cut off frequency [26] Gait events were subsequently defined and inverse dynamics analysis was performed in Matlab to calculate internal knee mome nts F or the purposes of this study, only the knee kinematics and kinetics are reported. T o account for variability in participant characteristics, such as height and body mass, we normalize d the internal knee moment [27] [28] Primary Outcome Variables Due to the design intention of the device being to actuate the knee the primary biomechanical outcome measures were restricted to the knee angle and internal knee moment in the device wearing leg. TBL peak torque and timing of the peak torque in terms of percent of the gait cycle was also reported. Normative knee angle has two characteristic peaks : a flexio n peak during loading response and a flexion peak during swing ( Figure 1 ) We investigate d these peaks as well as the knee angle value at initial contact, the maximal knee flexion during pre swing, and the maximal knee extension during mid stance. We also investigated both the peak knee flexion and extension during loading response to account for individuals who may land hyperextended. Normative internal knee moment has five characteristic peaks: a flexor peak during loading response, an extensor peak during loading response and mid stance, a flexor peak during terminal stance an extensor peak during pre swing and a flexor peak during late swing ( Figure 2 ) All five peaks were evaluated. The convention used was to express extension in knee angle and internal moment as positive [1], [27]

PAGE 10

Figure 1: An example of knee angle in normative gait Presented is the average and standard deviation of 13 steps of one control walking without the device. Figure 2 : An example of internal knee moment in normative gait. Presented is the average and standard deviation of 3 steps of one control walking without the device.

PAGE 11

Statistical Analysis We performed 2 way r epeated measures ANOVA s on knee angles and internal moments to address three questions. First, we investigated the differential effect of device condition (no device, free swing, default setting) by group (control vs. stroke). Second, we investigated the interaction of g roup and assistance percentage. Finally, we investigated the interaction of group and threshold Unpaired t performed to isolate effects in the presence of significant interactions and main effects. W hen data visualization was suggestiv e of effects e xploratory analyse s were conducted using one way ANOVAs within group and planned contrasts. Reported values in the results and discussion section s are constrained to the device wearing leg. All statistical tests were performed with JMP Pro 11.0.0 (SAS Institute Inc., Cary, NC), and a n Results What happened when the user put on the device? Stance : There were significant group x condition interactions observed in the knee angle at initial contact (F ( 2,52 ) = 4.83 p = 0.01 Figure 3 a ) and during loading r esponse ( Flexion : F (2, 52 ) = 5.85 p < 0.01 Figure 3 b ; Extension : F ( 2,51.4 ) = 4.72 p = 0.01 Figure 3 c ). There was a significant condition main effect during mid stance (F ( 2,51.5 ) = 6.42 p < 0.01 Figure 3 d ) There was a significant group x condition interaction observed in the knee angle at pre swing ( F ( 2,51.7 ) = 8.59 p < 0.01 ; Figure 3 e ). W e found no significant group x condition interactions in the internal knee moment at initial contact (F (2,39.2 ) = 0.4 5, p = 0.64, Figure 4 a ) during loading response ( LR: F (2, 36.7 ) = 1.25 p = 0.30, Figure 4b ), during mid stance ( Mst: F ( 2,38.5 ) = 0.28 p = 0.75 Figure 4 c ), or during combined loading response and mid stance ( LR & Mst: F ( 2,38.3 ) = 0.77 p = 0.47, Figure 4 d ) There were significant main effects observed in the internal knee moment during pre swing (Group: F ( 1,20.7 ) = 6.31 p = 0.02 ; Condition : F ( 2,39.4 ) = 3.35 p < 0.05 ; Figure 4 e ) Swing : We detected significant group x conditions interactions for knee angle in swing (F (2,25) = 13.3, p < 0.01 ; Figure 3 f ). TBL Torque: There were no significant group x condition interactions observed in the peak TBL torque ( F (1,25) = 1.54, p = 0.23 Figure 5 ), but there were significant group and condition main effects observed in the peak TBL torque (Condition: F (1,25) = 5.16, p = 0.03;

PAGE 12

Group: F (1,25) = 22.38, p < 0.01 Figure 5 ). For the control group the mean peak TBL torque in the default condition was 7.23 Nm (SD 5.7) occurring on average at 59.1% (SD 3 .1) of the gait cycle ( Figure 5 ) For the control group the mean peak TBL torque in the free swing condition was 6.09 Nm (SD 3.54) occurring on average at 59.6% (SD 3.24) of the gait cycle ( Figure 5 ) For the group pos t stroke the mean peak TBL torque in the default condition was 1.45 Nm (SD 1.66) occurring on average at 58.5% (SD 8.60) of the gait cycle ( Figure 5 ) For the group post stroke the mean peak TBL torque in the free swing condition was 1.11 Nm (SD 1.27) oc curring on average at 57.9% (SD 9.95) of the gait cycle ( Figure 5 )

PAGE 13

Figure 3: Condition effects on Knee Angle Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey. ND No device. FSw Free Swing. Significant group x condition interactions were observed at initial contact (a) and during loading response (b, c) pre swing (e) and swing (f) Knee flexion was increased at initial contact and during loading response in the group post st roke while wearing the device. Knee flexion was reduced during pre swing in the control group while wearing the device. d. A main effect of group was observed during mid stance. K nee extension during mid stance was significantly reduced when wearing the de vice.

PAGE 14

Figure 4 : Condition effect s on Knee Moment Subplo ts illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey. ND No device. FSw Free Swing. a d. There were no significant differences in peak internal knee moment at initial contact, or during loading response, mid stance, or loading response and mid stance. e. Peak internal knee extension moment during pre swing was significantly reduced while wear ing the device (F (1,39) = 6.56, p = 0.01).

PAGE 15

What happened when the assistance percentag e of the device was changed? Stance: There were no significant group x assistance percentage interactions found in the knee angle at initial contact (F (2,50) = 0.07 p = 0.93 Figure 6 a ), during loading response ( Flexion : F (2,51 ) = 0.14 p = 0.87 Figure 6 b ; Extension : F (2,50) = 0.06 p = 0.94 Figure 6 c ), during mid stance (F (2,50) = 0.38 p = 0.69 Figure 6 d ) or during pre swing (F ( 2,51 ) = 0.85 p = 0.43 Figure 6 e ) We found significant group main effects in the knee angle at initial contact ( F (1,31 ) = 7.45 p = 0.01 Figure 6 a ) and pre swing ( F (1,36 ) = 10.15 p < 0.01 Figure 6 e ). We found significant assistance percentage main effects in knee angle at loading response ( Extension: F (2,51 ) = 3.72 p =0.03 Figure 6 c ). W e found no significant group x assistance percentage interactions in the internal knee moment at initial contact (F (2,39 ) = 0.20 p = 0.82 Figure 7 a ), during loading response ( LR: F (2,38) = 0.05 p = 0.95, Figure 7b ), during mid stance (Mst: F (2,38) = 2.56 p = 0 .09 Figure 7 c ) during loading response and mid stance combined ( LR & Mst: F (2,38) = 1.67 p = 0.20, Figure 7d ) or during pre swing(F ( 2,39 ) = 1.21 p = 0.31 Figure 7 e ) Swing : We observed a significant assistance percentage main effect An assistance percentage of 70 resulted in significantly reduced knee flexion compared to an assistan ce percentage of 30 ( F ( 2,50 ) = 4.72 ; p = 0.01 Figure 6 f ). There were no significant group x Figure 5: TBL Torque during conditions Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey. FSw Free Swing. Peak torque was signifi cantly greater in the control group than in the group post stroke. Peak torque was significantly greater in the default condition than during the free swing condition, however the control group was still able to actuate the device in the free swing conditi on. The threshold percentage of device in the free swing condition was set to 95 percent. The timing of the peak torque is illustrated on the right for both groups, occurring late in pre swing on average.

PAGE 16

assistance percentage interactions during swing in the knee angle ( F (2,50) = 2.85 ; p = 0.07 Figure 6 f ). TBL Torque: There were significant group x condition interactions ob served in the peak TBL torque ( F (2,49 ) = 5.73 p < 0.01 Figure 8 ) The peak TBL torque at an assistance percentage of 50 and 70 were significantly greater than the peak TBL torque at an assistance percentage of 30 ( Figure 8 ) For the control group the mean peak TBL torque at an assistance percentage of 30 was 5.41 Nm (SD 3.49) occurring on average at 61 % (SD 2.9) of the gait cycle ( Figure 8 ) For the control group the mean peak TBL to rque at an assistance percentage of 50 was 7.23 Nm (SD 5.7) occurring on average at 59.1 % (SD 3.1) of the gait cycle ( Figure 8 ) For the control group the mean peak TBL torque at an assistance percentage of 70 was 7.06 Nm (SD 4.55) occurring on average at 59 % (SD 1.46) of the gait cycle ( Figure 8 ) The peak TBL torque did not differ across assistance percentages in the group post stroke ( Figure 8 ) For the group post stroke the mean peak TBL torque at an assistance percentage of 30 was 1.24 Nm (SD 1.51) occurring on average at 59 % (SD 8.99) of the gait cycle ( Figure 8 ) For the group post stroke the mean peak TBL torque at an assistance percentage of 50 was 1.45 Nm (SD 1.66) occurring on average at 58.6 % (SD 8.6) of the gait cycle ( Figure 8 ) For the group post stroke the mean peak TBL torque a t an assistance percentage of 7 0 was 1.28 Nm (SD 1.44) occurring on average at 57.6 % (SD 9.1) of the gait cycle ( Figure 8 )

PAGE 17

F igure 6 : Assistance effects on Knee Angle Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey. a. Knee flexion in the group post stroke was significantly greater than controls at initial contact. b. There was a main effect of group during loading response. Knee flexion was significantly greater at an assistance percentage of 30 than an assistance percentage of 50 during loading response. Knee flexion was not significantly different at an assistance percentage of 70 than an a ssistance percentage of 30 or 50. c d. There was no significant difference in knee angle during loading response or mid stance. e. Knee flexion was significantly greater for controls than for the group post stroke during pre swing. f. There was a main effe ct of group during swing. Knee flexion was significantly greater at an assistance percentage of 30 than an assistance percentage of 70 during swing

PAGE 18

Fi gure 7 : Assistance effects on Knee Moment Subplots illustrate means, standard error, and min to m ax. Control is illustrated in black. Post stroke is illustrated in grey. a e. There were no significant group x assistance percentage interactions or group or assistance percentage main effects in internal knee moment at any peaks in the gait cycle.

PAGE 19

What happened when the threshold percentage of the device was changed? Stance: There were no significant group x threshold percentage interactions found in the knee angle at initial contact (F (2,49 ) = 0.33 p = 0.72 Figure 9 a ), during loading response ( Flexion : F (2,48 ) = 0.30 p = 0.74 Figure 9 b ; Extension : F (2,48 ) = 1.11 p = 0.34 Fi gure 9 c ), during mid stance (F (2,48 ) = 0.75 p = 0.48 Figure 9 d ) or during pre swing ( F (2,48 ) = 0.06 p = 0.94 Figure 9 e ) There were significant group main effects in the knee angle at initial contact ( F (1,32 ) = 6.16 p = 0.02 Figure 9 a ) and during pre swing( F (1,32 ) = 7.32 p = 0.01 Figure 9 e ). W e found no significant group x threshold percentage interactions in the internal knee moment at initia l contact (F (2,35 ) = 0.13 p = 0.88 Figure 10 a ), during loading response (F (2,36 ) = 1.82 p = 0.18 Figure 10 b ), or during mid stance (F (2,36 ) = 2.61 p = 0.09 Figure 10 c ). There were significant device threshold main effects in the knee moment at initial contact ( F (2,35 ) = 3.36 p < 0.05 Figure 10 a ) There were significant group main effects in the peak knee moment at lo ading response and mid stance ( F (1,26 ) = 5.51 p = 0.03 Figure 10 d ) and during pre swing ( F (1 ,3 4) = 6.07 p = 0.02 Figure 10 e ) Figure 8: TBL torque and timing at different assistance percentages Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey. Peak torque was significantly greater in the control group t han in the group post stroke. Peak torque was significantly greater with an assistance percentage of 50 or 70 than with an assistance percentage of 30. The timing of the peak torque is illustrated on the right for both groups, occurring late in pre swing o n average.

PAGE 20

Swing : There were no significant group x threshold percentage interactions found in the knee angle during swing ( F (2,48 ) = 0.74 p = 0.48 Figure 9 f ), but there were significant group main effects in the in the knee angle during swing ( F (1,31 ) = 17.6 p < 0.01 Figure 9 f ) TBL Torque: There were no significant group x condition interactions ob served in the peak TBL torque ( F ( 2,50 ) = 0.68 p = 0.51 Figure 1 1 ), but there was a significant group main effect observed in the peak TBL torque (Group: F (1,29 ) = 16.24 p < 0.01 Figure 1 1 ). For the control group the mean peak TBL torque at a threshold percentage of 20 was 6.73 Nm (SD 4.51), occurring on average at 61.1 % (SD 2.91) of the gait cycle ( Figure 1 1 ) For the control group the mean peak TBL torque at a threshold percentage of 30 was 7.23 Nm (SD 5.70) occurring on average at 59.1 % (SD 3.06) of the gait cycle ( Figure 1 1 ) For the control group the mean peak TBL torque at a threshold percentage of 40 was 6.60 Nm (SD 4.91), occurring on average at 59.1 % (SD 2.99) of the gait cycle ( Figure 1 1 ) For the group post stroke the mean peak TBL torque at a threshold percentage of 20 was 1.57 Nm (SD 1.98), occurring on average at 58.4 % (SD 8.71) of the gait cycle ( Figure 1 1 ) For the group post stroke the mean peak TBL torque at a threshold percentage of 30 was 1.45 Nm (SD 1.66), occurring on average at 58.6% (SD 8.60) of the gait cycle ( Figure 1 1 ) For the group post stroke the mean peak TBL torque at a threshold percentage of 40 was 1.31 Nm (SD 1.37), occurring on average at 60.4 % (SD 9.54) of the gait cycle ( Figure 1 1 )

PAGE 21

F igure 9 : Threshold effects on Knee Angle Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is ill ustrated in grey. a) Knee flexion was significantly greater in the group post stroke than the control group at initial contact. e f) Knee flexion was significantly lower in the group post stroke during pre swing and swing. There were no significant group x threshold percentage interactions, and there were no significant threshold percentage main effects.

PAGE 22

F igure 10 : Threshold effects on Knee Moment Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey. a. Internal knee moment at initial contact at a threshold of 30 was significantly greater than a t a threshold of 20. Internal kn ee moment at a threshold of 40 was significantly greater than a threshold of 20 at initial contact. Internal knee moment was not significantly different from a threshold of 30 to 40. b c. There were no significant differences in internal knee moment at loa ding response and at mid stance. d. Peak internal knee extension moment in controls was significantly greater than the group post stroke. e. Peak internal knee extension moment was significantly greater in controls than in the group post stroke.

PAGE 23

Figure 1 2 : Reduced Hyperextension in the group Post Stroke. Many individuals of the group post stroke exhibited reduced knee extension during mid stance while wearing the device. Some of these individuals went from knee hyperextension to flexion. Fi gure 1 1 : TBL torque and timing at different assistance percentages Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey. Peak TBL torque was significantly greater in the control group. Changing the threshold percentage did not significantly change the torque or timing of the device.

PAGE 24

Discussion What happened when the user put on the device? We observed an increase in knee flexion during mid stance in both group s while wearing the device Although this may not be desired in the whole group, w e found that a subset of the group post stroke hyperextended their knee during mid stance without the dev ice and this was reduced when wearing the d evice ( Figure 1 2 ) This findin g may be particularly useful for clinicians treating individuals with hyperextension in stance [1] Knee flexion was also increased at initial contact and during loading response for the group post stroke. There was no significant difference in knee flexion in the control group at initial contact or during loading response Therefore during the majority of stance the device has a different effect on the biomechanics of the group post stroke than the control group. During the pre swing phase there is a significant reduction in flexion in the control group, but there is no significant difference in knee flexion when wearing the de vice for the group post stroke. I n norma l gait p eak knee angular velocity is reached during this period [27] so one possible explanation for this reduced flexion is that the device impedes achieving higher angular velocities during stance. Given that the walking patterns of individuals post stroke already exhibit reduced knee flexion in swing[1], it is reasonable to expect that this problem could be exacerbated by adding weight to the swing leg, however this was not the case. There was no significant difference in knee flexion during swing in the group post stroke but similar to the findings in pre swing discussed above w e did observe a reduction in knee flexion in the contro l group It is logical that a weight being added to the lower extremity could reduce knee flexion i n controls but seeing as knee flexion was reduced just prior during pre swing, the reduced knee flexion in swing cannot be simply attributed to the added load of the device We found that the device reaches peak actuation very late during the stance phase of the gait cycle Therefore, it would be necessa ry to improve the timing of the TBL actuation in future versions To provide extensor assistance in stance that is capable of producing a meaningful effect on internal knee moments, device actuation would need to be more precisely timed to occur by loading response and mid stance. Alternatively, t o positively influence knee flexion in swing following stroke, device actuation would need to be designed to assist knee flexion in swing.

PAGE 25

What happened when the assistance percentage of the device was changed? We found that changing the assistance percentage did not impact the internal knee moment during the stance phase in the group post stroke at any of the configurations tested ( Figure 7 ) Peak knee flexion was significantly greater during loading response at an assistance percentage of 30 than peak knee flexion at an assistance percentage of 50 as a group main effect but p eak knee flexion at an assistance percentage of 70 was not significantly different from peak knee flexion at an assistance percentage of 30 or 50 during loading response ( Figure 6 ) Therefore there is a non linear effect of assistance percentage on knee flexion during stance Loading response was also the only period in the gait cycle in which there were significa nt main effects of assistance percentage in kinematics or kinetics during stance During swing w e found that peak knee flexion at an assistance percentage of 30 was significantly reduced compared to an assistance percentage of 70. The intention of the device was to affect the stance phase, but this finding in swing means the device is causing undesirable effects when using a high assistance percentage. When using the current version of the device in therapy, clinicians would benefit from understanding that assistance percentages of 70 provide no advantage to lower assistance percentages. What happened when the threshold percentage of the device was changed? We found that changing the threshold percentage changed the internal knee moment at initial contact, but changing the threshold percentage did not impact the knee angle or internal knee moment throughout the rest of stance phase in the group post stroke at any of the configurations tested ( Figure 9 Fig ure 10 ) Li et al. 2015 found an increase in maximal knee extensor moment in participants of their training [9] The moment reported by Li was a global maximum wh ile we reported specific peaks so it is difficult to contrast these results directly I n norma l gait the global max imum knee moment occur s during loading response and mid stance ( Figure 1) [27] While we did not observe a cha nge in the peak moment for the group post stroke in this acute experiment we did find a significant increase in knee flexion during this period while wearing the device ( Figure 3 a c ) It is possible that over multiple sessions of training with t he device, the increased knee flexion we observed contributes to an improvement in knee extension moment [9]

PAGE 26

For the control group, our hypothesis that the device would reduce knee flexion during swing was confirmed but the hypothesis that the device would reduce peak knee extension moment during loading response and mid stance was not confirmed Despite testing various levels of assistance and threshold percentage o ur hypothesis that the group post stroke would experience a reduction in knee flexion in swing and a change in the internal knee moment during stance was not confirmed While experiencing no change in knee flexion during swing, upon wearing the device, the knee flexion of the group post stroke increased at initial contact and during loading response and this incr ease was independent of the configuration of the device. The torque output of the device at the tested assistance percentage s were low for the group post stroke compared to controls ( Figure 8 ) and c hanging the threshold percentage of the device with it s current programming did not increas e the torque output of the TBL for the group post stroke ( Figure 1 1 ) In the free swing condition the control group was able to activate the device despite a threshold of 95 ( Table 2 Figure 5 ) Therefore the device is capable of outputting larger torques, but the group post stroke was not able to tap into that capability. This information combined with the finding that the peak torque of the devices is reached late in the stance phase of the gait cycle or early into t he swing phase ( Figure 5, Figure 8, Figure 1 1 ) suggests that the sensing, timing, and programming of the device has room for improvement in order to maximize utility for the users post stroke. Specifically, the detection of loading through the foot sensor may need to be revised to accommodate for the altered loading pattern common in stroke [1] Limitations of the Study use of the device, however upo n inspection of the data, these participants did not deviate significantly from the mean of their respective groups for any of the variables analyzed. In conclusion a few valuable effects were observed regarding the Tibion PK100 bionic leg and its effec t on biomechanics of the knee First, wearing the device created no significant difference in peak knee extensor moment in an acute experiment. However, repeated exposure to i ncreased knee flexion during loading response in the group post stroke may provide an avenue to i ncrease knee extension moment post therapy Second, i n healthy controls, the device tended

PAGE 27

to reduce knee flexion in pre swing and swing, but it is not likely this effect can be primarily attributed to the mass of the device. Also, i n this acute experiment the device did not create any adverse effects in the group post stroke during the swing phase of the gait cycle. Finally after evaluating a range of device configurations, we found room for design improv ements in terms of timing and torque output W e observed that assistance percentages as high as 70 may not offer added benefit over lower assistance percentages ; this information could inform of the device in its current state We observed that changing the threshold percentage did not offer added benefit either, and in fact even at a threshold percentage of 20 the group post stroke was unable to significantly activate the device. If the device is intended to be marketed to assist therapy po st stroke, it is necessary to refine the sensing, timing, and programming to account for the unique gait pattern post stroke.

PAGE 28

References [1] Gait Posture vol. 4, no. 2, pp. 136 148, 1996. [2] E. J. Benjamin et al. 2018 Update: A Report From the American Heart Associa Circulation vol. 137, no. 12, p. e67 LP e492, Mar. 2018. [3] Quadriceps Muscle Fatigue on Stiff PLoS One vol. 9, no. 4, p. e94138, Apr. 2014. [4] Int. J. Precis. Eng. Manuf. vol. 16, no. 10, pp. 2219 2227, 2015. [5] A. J. Young and D of the art and Future Directions for Robotic Lower IEEE Trans. Neural Syst. Rehabil. Eng. vol. 25, no. 2, pp. 171 182, 2017. [6] robotic leg orthosis for rehabilitation and mobility enhan Proc. 31st Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. Eng. Futur. Biomed. EMBC 2009 vol. 94035, pp. 5030 5033, 2009. [7] current and emerging actuator techno Med. Eng. Phys. vol. 38, no. 4, pp. 317 325, Apr. 2016. [8] aided gait training in an individual with J. Neurol. Phys. Ther. vol. 36, no. 3, pp. 138 143, 201 2. [9] wearable robot Biomed. Mater. Eng. vol. 26, pp. S329 S340, 2015. [10] Am. J. Phys. Med. Rehabil. vol. 93, no. 11, pp. 987 994, Nov. 2014. [11] ionic knee orthosis in patients in a post stroke J. Med. Case Rep. vol. 6, p. 216, Jul. 2012. [12] A. M. Jette et al. evaluation of the disability component J. Gerontol. A. Biol. Sci. Med. Sci. vol. 57, no. 4, pp. M209 16, 2002. [13] S. M. Haley et al. J. Gerontol. A. Biol. Sci. Med. Sci. vol. 57, no. 4, pp. M217 22, Apr. 2002. [14] balance scale, fugl minute walk test in individuals with chronic stroke with different degrees of ankle Arch. Phys. Med. Rehabil. vol. 93, no. 7, pp. 1201 1208, 2012. [15] Stroke vol. 33, no. 7, pp. 1022 1027, 20 02. [16] the Fugl Meyer Assessment for Testing Motor

PAGE 29

Phys. Ther. vol. 73, pp. 447 454, 1993. [17] A. D. Pandyan, G. R. John son, C. I. Price, R. H. Curless, M. P. Barnes, and H. Rodgers, Clin. Rehabil. vol. 13, no. 5, pp. 373 383, Oct. 1999. [18] M. Blackburn, Phys Ther vol. 82, no. 1, pp. 25 34, 2002. [19] valuation Systems Test Phys. Ther. vol. 89, no. 5, pp. 484 498, May 2009. [20] techniques to improve the balance evaluation systems test: The mini J. Rehabil. Med. vol. 42, no. 4, pp. 323 331, 2010. [21] M. Godi, F. Franchignoni, M. Caligari, A. Giordano, A. M. Turcato, and A. Nardone, BESTest and Berg Phys. Ther. vol. 93, no. 2, pp. 15 8 167, Feb. 2013. [22] U. J. Rehabil. Med. vol. 37, no. 2, pp. 75 82, Mar. 2005. [23] Motion Di Available: https://c motion.com/v3dwiki/index.php/Digitizing_Pointer. [24] motion.com/v3dwiki/index.php/Main_Page. [25] [26] D. G. E. Robert J. Electromyogr. Kinesiol. vol. 13, no. 6, pp. 569 573, 2003. [27] D. A. Winter, The Biomechanics and Motor Control of Human Gait Second Edi. 1991. [28] A Gait Posture vol. 4, no. 3, pp. 222 223, 1996.

PAGE 30

Appendix A Illustration of Marker Set: Markers were placed on boney landmarks including: the spinous process of the 7 th cervical vertebrae (C7) the acromion (RSHO/LSHO) the lateral and medial epicondyles (RELA/RELB /LELA/LELB ) the radial and ulnar styloid process es (RWRA/RWRB/LWRA/LWRB) just below the head of the first, third, and fifth metacarpal (RMC#/LMC#) the iliac crest s (RASI/LASI) the lateral and medial epicondyles of the knees (RKNE/RKNM/LKNE/LKNM) the lateral and medial malleoli (RANK/RANM/LANK/LANM) the tip of the second metatarsal (R TIP /LTIP ) the calcaneus (at the same height as the second metatarsal head marker R HEE /L HEE ), the medial side of the first metatarsal head (RMT1/LMT1) and the lateral side of the fifth metatarsal head (RMT5/LMT5) Cluster placements were as follows: upper arm clusters in line with acromion and lateral epicondyle markers. Lower arm clusters in line with upper arm c lusters, thigh clusters in line with the greater trochanter and lateral knee marker, shank cluster in line with thigh clusters, foot clusters on top of shoes, and the pelvic cluster positioned directly over the posterior superior iliac spine (RPSI/LPSI)

PAGE 31

Appendix B Supplemental Plots Figure 3d Co n dition effect on knee angle during mid stance by group Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey.

PAGE 32

Figure 6 b : Assistance effect on knee angle during loading response by group Subplots illustrate means, standa rd error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey.

PAGE 33

Figure 6 f : Assistance effect on knee angle during swing by group Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey.

PAGE 34

Figure 10 a: Threshold effect on knee angle at initial contact by group Subplots illustrate means, standard error, and min to max. Control is illustrated in black. Post stroke is illustrated in grey.