Lower Limb Muscle Synergy During Daily Life Activities: A Way to Convey Intended Motions To a Robotic Assistive Device.

(1)

IN

DEGREE PROJECT MEDICAL ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2018,

Lower limb muscle synergy during daily life activities.

A way to convey intended motions to a robotic assistive device.

TERESA COLANGELO

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ENGINEERING SCIENCES IN CHEMISTRY, BIOTECHNOLOGY AND HEALTH

(2)

(3)

Lower limb muscle synergy during daily life activities.

A way to convey intended motions to a robotic assistive device.

TERESA COLANGELO

Degree Project in Medical Engineering Stockholm, Sweden 2018

Supervisor: Elena Gutierrez-Farewik Examiner: Dmitry Grishenkov

School of Engineering Sciences in Chemistry, Biotechnology and Health KTH Royal Institute of Technology

(4)

(5)

Abstract

Powered exoskeletons can assist patients suffering from motor dysfunctions. Recent researches are focused on how to improve the communication system between patient and device. Further research is needed in order to design an EMG based robotic assistive device able to convey intended motions to the patient. The primary need is the understanding of how EMG patterns from different muscles contribute to motions. Studies on muscle synergy have shown how different muscles of lower limbs contribute to gait. This study is aimed to expand the analysis to motions other than gait by analysing ten muscles around the right knee joint. The chosen muscle were soleus, gastrocnemius medialis, gastrocnemius lateralis, peroneus longus, tibialis anterior, rectus femoris, vastus medialis, vastus lateralis, biceps femoris and semitendinosus.

The main hypothesis is that specific movements are controlled by specific muscle synergies.

Motion data and EMG data of eight healthy subjects have been compared in order to outline a coordination pattern specific to four different movements: gait, gait stop and balance, sit to stand and stand to sit. Through the analysis of EMG signals, three muscle synergies have been identified including muscles from the same group, i.e. four plantar flexors, three quadriceps and two hamstrings. It was possible to conclude that the four movements were controlled by the same muscle synergies with different coordination patterns. Further research is recommended to expand the knowledge about muscle synergies.

(6)

(7)

Acknowledgements

I would like to thank my supervisor Elena Gutierrez-Farewik for giving me the possibility to choose this exciting project, for the passion she showed in teaching and for creating a stimulating and challenging environment in the research group.

I am deeply thankful to my family, for the immense support and understanding they always show, while being close or miles away. You always believe in me and make me stronger.

Thanks to all my friends for their true friendship and their constant support. A special thank you goes to the ones who participated in my experiment, you know who you are.

Thanks to ˚Asa Bartonek and Michael Remeringen at the Motoriklab at the Astrid Lindgren’s Barnsjukhus for the time they have spent teaching me.

(8)

(9)

List of Figures

2.1 Markers placement . . . 4

2.2 Reconstruction of motions in Vicon Nexus . . . 5

2.3 Schematic view of single muscle sinergies . . . 7

2.4 Schematic view of group muscle synergies . . . 7

3.1 Gait: EMG . . . 10

3.2 Gait: Normalized synergies . . . 11

3.3 Gait: Averaged group synergies . . . 12

3.4 Gait stop: EMG . . . 13

3.5 Gait stop: Normalized synergies . . . 14

3.6 Gait stop: Averaged synergies . . . 15

3.7 Sit to stand: EMG . . . 16

3.8 Sit to stand: Normalized synergies . . . 17

3.9 Sit to stand: Averaged synergies . . . 18

3.10 Stand to sit: EMG . . . 19

3.11 Stand to Sit: Normalized synergies . . . 20

3.12 Stand to sit: Averaged synergies . . . 21

A.1 Lokomat Pro . . . 34

A.2 GMFCS for children of age 6-12 . . . 36

A.3 Examples of orthoses . . . 38

B.1 Gait: Motions . . . 46

B.2 Gait stop: Motions . . . 47

B.3 Sit to stand: Motions . . . 48

B.4 Stand to sit: Motions . . . 49

C.1 Gait: Synergies . . . 52

C.2 Gait stop: Synergies . . . 53

C.3 Sit to stand: Synergies . . . 54

C.4 Stand to Sit: Synergies . . . 55

(12)

List of Tables

2.1 Anthropometric data . . . 3

2.2 Markers placement . . . 4

2.3 Movements . . . 4

2.4 Events . . . 6

2.5 Time normalization . . . 6

3.1 Synergies . . . 9

3.2 Gait: Muscles St.Dev. . . 10

3.3 Gait: Synergies St.Dev. . . 12

3.4 Gait stop: Muscles St.Dev. . . 13

3.5 Gait stop: Synergies St.Dev. . . 15

3.6 Sit to stand: Muscles St.Dev. . . 16

3.7 Sit to stand: Synergies St.Dev. . . 18

3.8 Stand to sit: Muscles St.Dev. . . 19

3.9 Stand to sit: Synergies St.Dev. . . 21

viii

(13)

List of Abbreviations

ADL activities of daily living ASIS anterior superior iliac spine BMI body mass index

CP cerebral palsy EMG electromyography

PSIS posterior superior iliac spine SCI spinal cord injury

SD standard deviation

(14)

(15)

Chapter 1

Introduction

Motor dysfunctions are one of the many consequences of the damage to the nervous system.

Cerebral palsy affects 2-3 out of 1000 children, causing severe lifelong motor impairments.^1,2 One of the main pathologies associated with cerebral palsy is crouched gait, a condition that worsens over time and negatively affects the quality of life. Traumatic spinal cord injuries happen to 1 out of 1000 people of all ages, mainly due to accidents related to traffic and sport.³ Damages to the spinal cord cause both temporary and permanent impairments to the motor system that can be treated with rehabilitation. Motor dysfunctions are also associated with stroke, a condition that affects 1.34% of the world’s population.⁴ Recent neurorehabilitation strategies aimed to treat motor dysfunctions have been focusing on active orthoses,⁵often called exoskeletons, as temporary or permanent assistive devices.^6–9 One of the biggest challenges of designing exoskeletons is to conceive a control system that is fast and intuitive for the patient.

Studies have been focusing on the improvement of communication systems between the device and the central nervous system, through the detection of EMG signals of the impaired limb.

The interpretation of EMG signals have been investigated by recent studies on muscle synergies aimed to describe coordination patterns that control human movement.^10–14 The majority of these studies is focused on gait, but the literature lacks research on other motions, e.g. sitting, standing, stair walking. When designing an assistive device aimed to support patients in their daily life activities, it is important that the support is extended to more than just walking assistance. This study is aimed to further investigate muscle synergies, in order to contribute to expand the knowledge on muscle coordination patterns.

1.1 Aim

The aim of this study is to understand how ten different muscles around the knee joint contribute to ADLs. The main purpose is to find the intent of motion through the analysis and comparison of EMG patterns and motion patterns.

1.2 Hypothesis

The hypothesis is that specific movements are controlled by specific muscle synergies and these coordination patterns can be used to predict the intent of the patient regarding specific movements such as gait, gait stop and balance, sit to stand and stand to sit.

(16)

(17)

Chapter 2

Methods

The chapter describes methods and procedures used in the study. An overview on the participants, the experimental protocol and the equipment used for data acquisition is given in section 2.1. Details on how the data were processed and analysed are given in section 2.2

2.1 Experimental setup and protocol

Participants Eight healthy subjects were recruited among students at the KTH Royal Insti- tute of Technology in Stockholm, Sweden. Anthropometric data were collected anonymously according to table 2.1. Mass and height were measured on the day of motion recordings.

The subjects were weighted while wearing markers, electrodes and EMG sensors, which added around 0.2 kg. The Body Mass Index (BMI) has been calculated as BM I =^{BodyM ass}_Height₂ [_m^kg₂].

Table 2.1: Anthropometric data of eight volunteers, represented as mean ± SD.

Feature Value

Age [years] 25.9±1.5

Sex [M:F] 4:4

Body Mass [kg] 66.7±9.1

Height [m] 1.7±0.1

Body Mass Index [kg/m²] 24.0±1.7 Physical activity [days/week] 3.7±2.2

Materials The Vicon Motion System was used to record motion data and EMG data. Mark- ers were placed according to figure 2.1. 23 markers were used according to table 2.2, to match the requirements of a modified Plug-in Gait model¹⁵ consisting of lower limbs and trunk.

(18)

2.1. EXPERIMENTAL SETUP AND PROTOCOL CHAPTER 2. METHODS

Figure 2.1: Schematic markers placement.¹⁶ Framed in red are the markers used in this study.

Table 2.2: Description of markers location.

Marker name Description

C7 7th cervical vertebra

T10 10th thoracic vertebra

CLAV Clavicle

STRN Sternum

RBAK Right back

LSHO/RSHO Left/Right shoulder LASI/RASI Left/Right ASIS LPSI/RPSI Left/Right PSIS LTHI/RTHI Left/Right thigh LKNE/RKNE Left/Right knee LTIB/RTIB Left/Right tibia LANK/RANK Left/Right ankle LHEE/RHEE Left/Right heel LTOE/RTOE Left/Right toe

Ten superficial electrodes were placed on the right leg, according to the indications of the Surface ElectroMyoGraphy for the Non-Invasive Assessment of Muscles project (SENIAM).¹⁷In this study the choice of muscles was limited by the number of electrodes, while previous studies analysed up to 31 different muscles in the lower leg.¹⁰The muscle chosen are soleus, gastrocnemius medialis, gastrocnemius lateralis, peroneus longus, tibialis anterior, rectus femoris, vastus medialis, vastus lateralis, biceps femoris and semitendinosus.

Motion data and EMG data were collected and preprocessed using the softwares Vicon Nexus 2.6¹⁸and ProEMG.¹⁹The results were then further processed and plotted in MATLAB.

Procedure Subjects were asked to perform four different motions several times before recording, in order to become comfortable and familiar with the procedure. Description on how the movements were performed is given in table 2.3.

Table 2.3: Movements analysed in the study with description on how they were performed.

Movement Description

Gait The subjects were asked to walk naturally and try to place the right foot on Force Plate 1 and the left foot on Force Plate 2 (Figure 2.2a).

Gait stop and balance The subjects were asked to stop the walking naturally when the right foot was on Force Plate 1 (Figure 2.2b) and balance for ∼2.5 seconds.

Sit to stand The subjects were asked to naturally sit on a stool (45cm height) and rest, without placing their hands on their legs (Figure 2.2c).

Stand to sit The subjects were asked to naturally stand from a stool (45cm height) and balance, without placing their hands on their legs (Figure 2.2d).

4

(19)

2.2. DATA PROCESSING AND ANALYSIS CHAPTER 2. METHODS

(a) Gait. (b) Gait stop and balance.

(c) Sit to stand. (d) Stand to sit.

Figure 2.2: Reconstruction of motions in Vicon Nexus. Green segments indicate right side, red segments indicate left side, blue segments indicate trunk and pelvis. (a) Force plate 1 (green) shows foot strike of the right leg, while force plate 2 (red) relates to the left foot strike. (b) Force plate 2 remained inactivated due to balance of left foot outside the force plate. (c),(d) Force plate 1 relates to the left side (red) while force plate 2 relates to the right side (green).

2.2 Data processing and analysis

Plug-in Gait Dynamic Model Movement acquisitions were preprocessed in Nexus using the Plug-in Gait Dynamic Model²⁰ to obtain joint’s angles, moments and powers. The model requires information on all markers of lower limbs and pelvis. Information on missing markers were corrected with automatic and manual gap filling, necessary in order to guarantee accuracy during the calculation of angles, moments and powers. This topic is further discussed in sections 4.1.4 and 4.2.

Events generation Events were generated in Nexus both automatically (when informations about the force plates were given) and manually. Manual generation is based on visual changes of the reconstructed model in Nexus, which relies on the quality of the data acquired and the

(20)

experience of the user. Table 2.4 shows information of significant events for the four movements.

Table 2.4: Description of significant events of four movements. Manually generated events are indicated with M.

Movement Event name Description

Gait R1heel First right heel strike Ltoe [M] Left toe off

Lheel Left heel strike Rtoe [M] Right toe off

R2_heel Second right heel strike Gait stop Rheel Right heel strike

Ltoe [M] Left toe off Lheel [M] Left heel strike

Sit to stand S [M] Standing (the moment when the body leaves the stool) Legext [M] Full leg extension (end of standing/start of balancing) Stand to sit Legf lex [M] Start of leg flexion

S [M] Sitting (the moment when the body touches the stool)

Time normalization Time normalization was necessary to compare data of different subjects. Since gait analysis is a well known process and the Plug-in Gait Model provides automatic time normalization, the convention used for gait was the same used in literature.^6,10,13,21 Not enough studies on movements other than gait have yet been performed, hence literature lacks a convention for time normalization of gait stop and balance, sit to stand and stand to sit.

For this reason the choice of each movement’s cycle has been arbitrary and based on how the motions were performed. Table 2.5 shows the starting/ending events used for normalization and events considered significant for the analysis.

Table 2.5: Description of events used for time normalization. Full descriptions of events are shown in table 2.4

Movement Start End Significant event

Gait R1heel R2heel Rtoe

Gait stop and balance Rheel ∼2.5s after Lheel Ltoe, Lheel

Sit to stand ∼0.5s before S t^∗ after Leg_ext S, Leg_ext Stand to sit ∼0.5s before Legf lex ∼1.5s after S Legf lex,S t^∗= t_Leg_ext− t_S

EMG preprocessing The Myon proEMG¹⁹ software was used to process EMG signals di- rectly from Nexus. The signals were filtered (Butterworth 10-200Hz), rectified, smoothed with RMS at 45ms and moving average 40ms, normalized in time and in amplitude at the maximum peak.

Data extraction Vicon Nexus provides motion data in files with extension .c3d, containing informations on marker position, joint’s angles, powers, moments, ground reaction forces and EMG data. These informations were extracted with the c3d2OpenSim MATLAB extraction tool²² modified to match the user’s requirements.

Muscle analysis - Averaging In order to compare EMG data of eight different subjects, signals of the same muscle were averaged and plotted in MATLAB. Mean values and standard deviations were then calculated to quantify the differences among subjects.

6

(21)

Muscle analysis - Synergies Muscle synergies have been identified among muscles of the same functional group. Soleus, gastrocnemius medialis and gastrocnemius lateralis are the biggest contributors of plantar flexion, also regulated by peroneus longus. Peroneus longus and tibialis anterior are antagonist muscles, acting to control and oppose each other’s activity.

However, these two muscles have not been compared due to a lack of pattern that could relate the two. Rectus femoris, vastus medialis and vastus lateralis are three of the quadriceps, responsible for knee extension and hip flexion. Biceps femoris and semitendinosus are two of the hamstrings, responsible for knee flexion and hip extension. To quantify the relationship between muscles and define muscle synergies, each group of muscles has been first averaged across single subjects, then mean values and standard deviations have been calculated and analysed (see figure 2.3 for schematic representation). Finally, further averaging have been performed including all subjects into one calculation (see figure 2.4 for schematic representation).

S1 S2 S3 S4 S5 S6 S7 S8

Figure 2.3: Schematic view of averaging procedure and standard deviation calculations for single muscle synergies for each subject.

Figure 2.4: Schematic view of averaging procedure and standard deviation calculations for group muscle synergies. Each muscle (e.g. SOLEUS) is the average of the eight correspondent muscles for each subject (soleus of S1, soleus of S2 etc.). Each synergy (e.g. M1) is the average of correspondent averaged muscle (e.g. M1 includesSOLEUS,GASTROCNEMIUS MEDIALIS,GASTROCNEMIUS LATERALIS, PERONEUS LONGUS).

(22)

(23)

Chapter 3

Results

This chapter is divided into four sections showing the results for the four movements. In each section the results of muscle activity are presented with plots of each muscle averaged within the group (Figures 3.1, 3.4, 3.7, 3.10) followed by a corresponding table showing values of standard deviation (Tables 3.2, 3.4, 3.6, 3.8). Results from muscle synergy’s analysis are then presented with plots showing three synergies for each subject (Figures 3.2, 3.5, 3.8, 3.11), a table sum- marizing the values of corresponding standard deviations (Tables 3.3, 3.5, 3.7, 3.9) and finally plots of group averaged synergies (Figures 3.3, 3.6, 3.9, 3.12). The results of muscle activity are presented as normalized amplitude of EMG signals to peak values. Since the value of standard deviation resulted to be rather high and no similar methods of analysis were previously used, a value of 15% standard deviation was considered an indicator for significance. The indication in percentage refers to the values of normalized amplitude of EMG signals, also expressed in percentage of peak value. The following muscle synergies have been identified and arbitrarily named, as shown in table 3.1.

Table 3.1: Three muscle synergies and corresponding muscles.

Synergy Muscles included

M1 Soleus, gastrocnemius medialis, gastrocnemius lateralis, peroneus longus M2 Rectus femoris, vastus medialis, vastus lateralis

M3 Biceps femoris, semitendinosus

Plots of motions and EMG signals without amplitude normalization are shown in appendix B and C respectively.

(24)

3.1. GAIT CHAPTER 3. RESULTS

3.1 Gait

Figure 3.1: Gait: Averaged muscle activity of ten muscles of the right leg across subjects.

Black lines indicates the mean values of EMG, blue line shadings indicate the range defined as mean±SD. The amplitude of each EMG signal has been normalized to peak value across single trials. Vertical gray lines and shadings indicate mean±SD of right toe off.

Table 3.2: Gait : Mean standard deviation of averaged EMG signals of single muscles among subjects. Values are expressed as percentage of averaged peak value. Values > 15% are indicated with *.

Muscle SD [%]

Peroneus longus 14.7 Tibialis anterior 16.6*

Soleus 14.6

Gastrocnemius medialis 10.5 Gastrocnemius lateralis 9.0 Vastus medialis 15.7*

Vastus lateralis 13.8 Rectus femoris 16.2*

Semitendinosus 12.4 Biceps femoris 15.1*

10

(25)

M1 M2 M3

S1

S2

S3

S4

S5

S6

S7

S8

Figure 3.2: Gait: Identification of three muscles synergies (columns M1, M2, M3) for each subject (rows S1-S8). M1 includes soleus (dark blue), gastrocnemius medialis (light blue), gastrocnemius lateralis (cyan) and peroneus longus (purple). M2 includes rectus femoris (dark green), vastus medialis (light green) and vastus lateralis (yellow). M3 includes biceps femoris (dark yellow) and semitendinosus (red). Black lines indicate M1, M2, M3 as the mean value of the corresponding muscles, yellow shadings indicate the range of mean±SD. Vertical gray lines correspond to right toe off. Amplitude of EMG signals has been normalized to peak value across single trials.

(26)

Table 3.3: Gait : Standard deviation of muscle synergies for each subject, expressed as percentage of normalized amplitude. Total values express the average of standard deviation for each muscle synergy. Values > 15% are indicated with *.

Subject M1 SD [%] M2 SD [%] M3 SD [%]

S1 10.4 11.5 11.4

S2 7.5 16.0* 4.8

S3 7.8 15.1* 3.6

S4 10.4 14.0 4.6

S5 12.6 18.1* 6.8

S6 5.8 10.6 9.7

S7 19.1* 16.5* 5.1

S8 6.3 8.5 7.2

All subjects 8.4 10.3 12.4

(a) (b) (c)

Figure 3.3: Gait: Averaged group synergies, normalized in amplitude to average peak value.

(a) M1 includes averaged soleus (blue), gastrocnemius medialis (light blue), gastrocnemius lateralis (cyan) and peroneus longus (purple). (b) M2 includes rectus femoris (dark green), vastus medialis (light green) and vastus lateralis (yellow). (c) M3 includes biceps femoris (dark yellow) and semitendinosus (red). Black lines indicate M1, M2, M3 as the mean value of the corresponding muscles. Blue shadings indicate the range calculated as mean±SD. Vertical gray line and shading indicate mean±SD of right toe off.

12

(27)

3.2. GAIT-STOP AND BALANCE CHAPTER 3. RESULTS

3.2 Gait-stop and balance

Figure 3.4: Gait stop and balance: Averaged muscle activity of ten muscles of the right leg across subjects. Black lines indicate the mean values of EMG, blue line shadings indicate the range defined as mean±SD. The amplitude of each EMG signal has been normalized to peak value across single trials, time normalization has been calculated between right heel strike and

∼2.5 seconds after left heel strike. Vertical gray lines and shadings indicate mean±SD of left toe off ( ) and left heel strike ( ).

Table 3.4: Gait stop and balance: Standard deviation of averaged EMG signals of single muscles among subjects. Values are expressed as percentage of averaged peak value. Values > 15% are indicated with *.

Muscle SD [%]

Peroneus longus 19.3*

Tibialis anterior 9.9

Soleus 16.3*

Gastrocnemius medialis 17.8*

Gastrocnemius lateralis 20.6*

Vastus medialis 17.3*

Vastus lateralis 19.7*

Rectus femoris 19.0*

Semitendinosus 15.1*

Biceps femoris 17.9*

(28)

M1 M2 M3

S1

S2

S3

S4

S5

S6

S7

S8

Figure 3.5: Gait stop and balance: Identification of three muscles synergies (columns M1, M2, M3) for each subject (rows S1-S8). M1 includes soleus (dark blue), gastrocnemius medialis (light blue), gastrocnemius lateralis (cyan) and peroneus longus (purple). M2 includes rectus femoris (dark green), vastus medialis (light green) and vastus lateralis (yellow). M3 includes biceps femoris (dark yellow) and semitendinosus (red). Black lines indicate M1, M2, M3 as the mean value of the corresponding muscles, yellow shadings indicate the range of mean±SD. Vertical gray lines and shadings indicate mean±SD of left toe off ( ) and left heel strike ( ). Amplitude of EMG signals has been normalized to peak value across single trials, time normalization has been calculated between right heel strike and ∼2.5 seconds after left heel strike.

14

(29)

Table 3.5: Gait stop and balance: Standard deviation of muscle synergies for each subject, expressed as percentage of normalized amplitude. Total values express the average of standard deviation for each muscle synergy. Values > 15% are indicated with *.

S1 17.4* 10.7 6.6

S2 10.4 4.6 9.7

S3 11.6 10.7 4.7

S4 20.6* 13.3 8.7

S5 16.6* 17.0* 15.3*

S6 12.9 10.4 7.9

S7 13.2 9.1 3.5

S8 11.2 9.6 3.5

All subjects 13.6 16.1* 15.5*

(a) (b) (c)

Figure 3.6: Gait stop and balance: Averaged group synergies, normalized in amplitude to average peak value, normalized in time between right heel strike and ∼2.5 seconds after left heel strike. (a) M1 includes averaged soleus (blue), gastrocnemius medialis (light blue), gastrocnemius lateralis (cyan) and peroneus longus (purple). (b) M2 includes rectus femoris (dark green), vastus medialis (light green) and vastus lateralis (yellow). (c) M3 includes biceps femoris (dark yellow) and semitendinosus (red). Black lines indicate M1, M2, M3 as the mean value of the corresponding muscles. Blue shadings indicate the range calculated as mean±SD. Vertical gray lines and shadings indicate mean±SD of left toe off ( ) and left heel strike ( ).

(30)

3.3. SIT-TO-STAND CHAPTER 3. RESULTS

3.3 Sit-to-stand

Figure 3.7: Sit to stand: Averaged muscle activity of ten muscles of the right leg across subjects.

Black lines indicate the mean values of EMG, blue line shadings indicate the range defined as mean±SD. The amplitude of each EMG signal has been normalized to peak value across single trials, time normalization has been calculated between ∼0.5 seconds before standing up and t^∗ seconds after leg extension (t^∗ = t_circle− ttriangle). Vertical gray lines and shadings indicate mean±SD of standing ( ) and leg extension ( ).

Table 3.6: Sit to stand : Standard deviation of averaged muscles among subjects. Values are expressed as percentage of averaged peak value. Values > 15% are indicated with *.

Muscle SD [%]

Tibialis anterior 11.9

Soleus 16.1*

Vastus medialis 12.3 Vastus lateralis 15.2*

Rectus femoris 11.5 Semitendinosus 14.2 Biceps femoris 15.0

16

(31)

M1 M2 M3

S1

S2

S3

S4

S5

S6

S7

S8

Figure 3.8: Sit to stand: Identification of three muscles synergies (columns M1, M2, M3) for each subject (rows S1-S8). M1 includes soleus (dark blue), gastrocnemius medialis (light blue), gastrocnemius lateralis (cyan) and peroneus longus (purple). M2 includes rectus femoris (dark green), vastus medialis (light green) and vastus lateralis (yellow). M3 includes biceps femoris (dark yellow) and semitendinosus (red). Black lines indicate M1, M2, M3 as the mean value of the corresponding muscles, yellow shadings indicate the range of mean±SD. Vertical gray lines and shadings indicate mean±SD of standing ( ) and leg extension ( ). Amplitude of EMG signals has been normalized to peak value across single trials, time normalization has been calculated between ∼0.5 seconds before standing up and t^∗ seconds after leg extension (t^∗= tcircle− ttriangle).

(32)

Table 3.7: Sit to stand : Standard deviation of muscle synergies for each subject, expressed as percentage of normalized amplitude. Total values express the average of standard deviation for each muscle synergy. Values > 15% are indicated with *.

S1 5.7 4.8 4.9

S2 16.4* 1.4 2.4

S3 12.5 3.9 7.2

S4 12.2 5.7 11.4

S5 12.3 9.8 6.8

S6 15.2* 5.3 8.7

S7 19.1* 6.4 5.4

S8 9.2 9.2 3.6

(a) (b) (c)

Figure 3.9: Sit to stand: Averaged group synergies, normalized in amplitude to average peak value, normalized in time between ∼0.5 seconds before standing up and t^∗ seconds after leg extension (t^∗ = t_circle − t_triangle). (a) M1 includes averaged soleus (blue), gastrocnemius medialis (light blue), gastrocnemius lateralis (cyan) and peroneus longus (purple). (b) M2 includes rectus femoris (dark green), vastus medialis (light green) and vastus lateralis(yellow).

(c) M3 includes biceps femoris (dark yellow) and semitendinosus (red). Black lines indicate M1,M2,M3 as the mean value of the corresponding muscles. Blue shadings indicate the range calculated as mean±SD. Vertical gray lines and shadings indicate mean±SD of standing ( ) and leg extension ( ).

18

(33)

3.4. STAND-TO-SIT CHAPTER 3. RESULTS

3.4 Stand-to-sit

Figure 3.10: Stand to sit: Averaged muscle activity of ten muscles of the right leg across subjects. Black lines indicate the mean values of EMG, blue line shadings indicate the range defined as mean±SD. The amplitude of each EMG signal has been normalized to peak value across single trials, time normalization has been calculated between ∼0.5 seconds before leg flexing and ∼1.5 seconds after sitting. Vertical gray lines and shadings indicate mean±SD of standing ( ) and leg extension ( ).

Table 3.8: Stand to sit : Standard deviation of averaged muscles among subjects. Values are expressed as percentage of averaged peak value. Values > 15% are indicated with *.

Muscle SD [%]

Tibialis anterior 15.9*

Soleus 15.2*

Vastus medialis 10.8 Vastus lateralis 11.9 Rectus femoris 12.7 Semitendinosus 14.6 Biceps femoris 12.3

(34)

M1 M2 M3

S1

S2

S3

S4

S5

S6

S7

S8

Figure 3.11: Stand to Sit: Identification of three muscles synergies (columns M1, M2, M3) for each subject (rows S1-S8). M1 includes soleus (dark blue), gastrocnemius medialis (light blue), gastrocnemius lateralis (cyan) and peroneus longus (purple). M2 includes rectus femoris (dark green), vastus medialis (light green) and vastus lateralis (yellow). M3 includes biceps femoris (dark yellow) and semitendinosus (red). Black lines indicate M1, M2, M3 as the mean value of the corresponding muscles, yellow shadings indicate the range of mean±SD. Vertical gray lines and shadings indicate mean±SD of leg flexion ( ) and sitting ( ). Amplitude of EMG signals has been normalized to peak value across single trials, time normalization has been calculated between ∼0.5 seconds before leg flexing and ∼1.5 seconds after sitting.

20

(35)

Table 3.9: Stand to sit : Standard deviation of muscle synergies for each subject, expressed as percentage of normalized amplitude. Total values express the average of standard deviation for each muscle synergy. Values > 15% are indicated with *.

S1 10.2 1.7 2.6

S2 18.6* 2.2 2.7

S3 11.7 3.9 4.9

S4 13.7 9.2 8.5

S5 8.6 7.2 10.8

S6 14.3 4.8 3.1

S7 17.0* 4.8 2.3

S8 7.0 5.1 8.3

(a) (b) (c)

Figure 3.12: Stand to sit: Averaged group synergies, normalized in amplitude to average peak value, normalized in time between ∼0.5 seconds before leg flexing and ∼1.5 seconds after sitting.

(a) M1 includes averaged soleus (blue), gastrocnemius medialis (light blue), gastrocnemius lateralis (cyan) and peroneus longus (purple). (b) M2 includes rectus femoris (dark green), vastus medialis (light green) and vastus lateralis(yellow). (c) M3 includes biceps femoris (dark yellow) and semitendinosus (red). Black lines indicate M1,M2,M3 as the mean value of the corresponding muscles. Blue shadings indicate the range calculated as mean±SD. Vertical gray lines and shadings indicate mean±SD of leg flexion ( ) and sitting ( ).

(36)

(37)

Chapter 4

Discussion

The main findings are discussed at the beginning of this chapter, based on the results showed in chapter 3. An overview of the biggest limitations observed in this study is given in section 4.2. Finally, some suggestions on future studies and their importance are given.

4.1 Main Findings

The primary hypothesis of this study was that specific movements are controlled by specific muscle synergies. This hypothesis is confirmed by the main findings which show how muscles from the same group present the same EMG pattern for the same motion of different subjects.

4.1.1 Gait

For gait analysis the averaging process of EMG signals and muscle synergies showed significant results for all three muscle synergies, as shown by table 3.3 and figures 3.2 and 3.3. M1 contributes to gait mainly during the terminal stance phase (25-50% of gait cycle) when the left leg is brought forward, showing that plantar flexors help the body balancing on one foot, dropping their activity as soon as the left foot strikes the ground. The coordination of the four muscle of M1 shows remarkable timing and pattern for at least six of the eight subjects.

The biggest discrepancy is shown by subject S7 and a clear difference can also be seen for subject S5. The two subjects participated in the second and first session respectively, therefore the inexperience of the investigator may have influenced the electrodes’ placement, especially the one recording peroneus longus in subject S7. The placement of peroneus longus’ electrode affected also the results for subject S1, as pointed out further in the discussion. However, the pattern of M1 in the second half of the gait cycle is more consistent with the other subjects, suggesting the validity of the results.

The contribution of M3 to gait is less clear compared to M1 but a general pattern can still be observed. All subjects show initial burst of activity during loading response (0-10% of gait cycle) and the first half of midstance (10-20%) and a final burst towards the end before the second right heel strike. The biggest discrepancy is shown by subjects S1, S5 and S6 whose plots present a significant activity in the middle of the cycle. In subject S1 some discrepancies are observed internally, as can be seen by the yellow shading in figure 3.2. A possible reason behind this difference is the electrode’s misplacement due to higher body fat, in fact subjects S1 and S6 showed the two highest BMIs, while S5 was the first subject tested.

Plots for M2 showed high variability both for single subjects and for the group. Therefore it is not possible to define M2 as a muscle synergy for gait, even though the muscles are all of the same group, i.e. quadriceps. This issue is further discussed in section 4.2.

The results for M1 and M3 are consistent with the ones from healthy subjects in the study by Clark et al. (2009).¹³A total of four modules were described and it is possible to observe a conformity between M1 of this study and module C2 in Clark’s, which shows that soleus and gastrocnemius medialis are the dominant contributors in the module and, moreover, they follow the same coordination pattern as M1. Module C4 is dominated by the activity of lateral and medial hamstrings, showing a similar pattern to M3.

(38)

4.1. MAIN FINDINGS CHAPTER 4. DISCUSSION

4.1.2 Gait-stop and balance

As can be seen in table 3.5 and figures 3.5 and 3.6, high values of standard deviations limit the significance of the analysis. One of the biggest discrepancies concerns subject S5. The time of balancing for this subject was shorter compared to other subjects, therefore time normalization and amplitude normalization of the whole group have been affected. However, similar patterns can still be identified.

M1 has its peak activity during the first 25% of the gait stop cycle, between right heel strike and left heel strike. Subjects S2, S3, S6, S7 and S8 present similar patterns. Peroneus longus in subject S1 affects dramatically the pattern of M1, as previously mentioned. Muscle activity decreases below 20% during balancing phase suggesting double support from both legs. However, subjects S2 and S7 present bursts of muscle activity during balancing phase, suggesting that they were shifting the weight between left and right leg while standing.

The pattern of M2 looks similar for all subjects but S8, starting with a peak of activity in the loading response (first 5% of gait stop cycle), decreasing in the stance phase (5-25%) and slightly bursting again at the moment of left heel strike. During balance the muscles set to constant values. Subject S5 also presents the same pattern with two peaks of activity in the loading response and left heel strike. Even though it is not possible to observe the muscle activity levelling off during balancing phase, a decreasing pattern after left heel strike suggest that the muscle activity might level off to lower level for subject S5 as for the other subjects.

Subject S4 presents significant irregularities of vastus lateralis (Fig. 3.5, yellow), suggesting electrode’s displacement. The rules on how to place the electrodes on vastus lateralis given by the SENIAM project¹⁷ and showed in table 2.2 require a certain level of experience, lack of which can lead to loss of accuracy.

M3 shows a significant initial peak during loading response for all subjects but S4. During stance phase the muscle activity drops down to 20% and below, and shows a peak of activity at left heel strike or slightly after. It is important to point out that the event generation was made manually for left toe off (Fig. 3.5, gray square) and left heel strike (Fig. 3.5, gray diamond), resulting in low accuracy. For this reason, assuming that a more accurate event generation system would improve the results, it can be concluded that the coordination patterns for M3 are similar among subjects.

A study performed on healthy subjects by Torres-Oviedo and Ting (2007)¹¹ analysed the contribution of muscle synergies to posture. Their methods consisted in observing muscle activation during posture described as follows; quiet standing followed by a ramp-and-hold pertur- bation followed by balance. This difference in methodology does not allow a direct comparison between synergy patterns of M1, M2, M3 and the six synergies observed by Torres-Oviedo and Ting. However, their results confirm the validity of choosing muscle from the same functional group for posture and balance. In their findings, synergy 1 includes soleus, peroneus longus, gastrocnemius medialis and lateralis, i.e. the same muscles of M1. In synergy 2 the main contributors are tensor fascia lata, rectus femoris, gluteus medius, vastus medialis and lateralis, which partially mirrors the activity of M2. Finally, synergy 3 consist on semitendinosus, semimembranosus and biceps femoris long head, mirroring M3.

4.1.3 Sit-to-stand

The results for M1 showed in figures 3.8 and 3.9 and the high variability of standard deviations indicated in table 3.6 do not allow to observe a coordination pattern for sit to stand. The reason behind the high discrepancy can be connected to the way the subjects performed the movement. They were asked to perform the movements as naturally as possible and as a result, almost all subjects confirmed to have slightly lifted their heels off the ground, balancing on their ball and toes for fractions of a second, affecting unpredictably the activity of plantar flexors.

The coordination patterns showed by M2 are visually very clear for all subjects and its validity is mirrored by values of standard deviations < 15% (Tab. 3.7). The three quadriceps have their peak activity right after the body leaves the stool, then the activity decreases rapidly during the standing up phase and levels off a level below 10%. The events of sit to stand have been manually generated, therefore their accuracy have been affected, as can be seen for subjects S3 and S4.

The contribution of M3 to sit to stand presents similar patterns for all subjects, with only few

24

(39)

4.2. LIMITATIONS CHAPTER 4. DISCUSSION

differences. The peak activity appears between the moment of standing up from the stool and the end of the standing up phase. The similarity of muscle activity between M2 (knee extensors) and M3 (knee flexors) is consistent with the occurrence of muscle coactivation, i.e. antagonist muscle simultaneous contraction aimed to stabilize the joint and control the movement. They both have their peak at the beginning of standing, when the knee is flexed at 40 degrees.

Subjects S1, S3, S4, S5 show some differences in M3 after the standing up phase, when the body is standing in balancing position, suggesting that muscles were working to maintain knee extension.

4.1.4 Stand-to-sit

As for sit to stand, the coordination patterns for M1 are unclear due to subjects lifting up their heels. M1 behaves differently for each subject as shown in figures 3.11 and 3.12 and by the high values of standard deviations of all plantar flexors (Tab. 3.8).

M2 shows a clear pattern described by a rapid increase of muscle activity starting at leg flexion (Fig. 3.11, gray triangle) which peaks before the moment when the subjects sit on the stool (Fig. 3.11, gray circle). The manual generation of events has certainly caused a big discrepancy for subject S5 and a slight one for subject S4, whose M2 has clearly the same pattern but results misplaced compared to the events of reference. Stand to sit and sit to stand movements had the lowest quality of reconstruction in Nexus.

As already observed for the sit to stand movement, M3 and M2 present similar patterns due to muscle coactivation. The initial activation for subjects S3, S4 and S5 differs from the one of the others due to differences during the standing position, as already observed for the sit to stand movement. Both M2 and M3 have a rapid increase in muscle activity, depending on the degree of flexion of the knee joint, reaching the peak when the knee is flexed at 50 degrees.

4.2 Limitations

One of the biggest limitations of this study is the high variability showed by the values of standard deviations. This limited the individuation of muscle synergies among muscles from the same functional group, i.e. plantar flexors, quadriceps and hamstrings. To be able to observe muscle synergies of muscles from different muscle groups, statistical tools as ANOVA analysis can be used and has shown interesting results.^10,12Statistical analysis was not possible due to the reduced number of repetitions for each subject. In fact, only one repetition for each movement was recorded for each subject, hence the analysis was limited to statistical tools as average and standard deviation.

Another limitation relates to the reduced number of EMG sensors available. The activity of only ten muscles was recorded, only on the right leg. More electrodes are needed to investigate both legs and more muscles, hence the analysis can be a more relevant description of how movements are really performed.

The use of superficial electrodes introduced important sources of error, due to the presence of body fat between the electrode and the muscle’ surface. This can be partially solved during filtering and smoothing, but important informations might be lost. Correct electrodes placement is also limited by the investigator’s experience, limitation that falls under human error.

This type of error is also occurring during the process of event generation in Vicon Nexus.

Automatic detection of event is possible under certain conditions (force plate event detection or percentage of markers in specific location), but many events were generated manually by the investigator, reducing precision and accuracy.

Motion capture systems as the Vicon rely on capturing markers’ positions. During data acquisition, markers can be hidden due to the type of motion performed. Sit to stand and stand to sit motions, when performed as naturally as possible, required bending the trunk over to balance the shift of the center of mass. This often resulted in hiding important markers as RASI and LASI, essential to calculate angles, moments and powers. This aspect could be corrected only partially during post processing in Nexus.

(40)

4.3. FUTURE WORK AND POSSIBLE APPLICATIONS CHAPTER 4. DISCUSSION

4.3 Future work and possible applications

As already mentioned, few studies have been performed on ADLs. This study showed that important informations about movements can be retrieved by the analysis of muscle synergies.

Studies on gait are numerous but it is clear that more needs to be investigated about motions other than gait. In this study an attempt to investigate motions as sit to stand and stand to sit has been performed. Movements like sit to stand, stand to sit, stair walking and walking on slope are only few of the movements of ADLs that need to be studied. The analysis of a wider range of muscle is also suggested, to have a more complete overview of how human motion is performed.

The study of muscle synergies can lead to the design of intent-driven EMG-based robotic assistive devices which can help several patients with motor dysfunctions. For this reason, it is necessary to investigate not only healthy subjects but also patients suffering from muscle weaknesses and other motor dysfunctions and how they adapt to perform ADLs.

26

(41)

Chapter 5

Conclusion

The contribution of different muscles to four different movements have been investigated for eight healthy subjects, in order to establish how different muscles contribute to different movements of ADLs. Two muscle synergies have been identified for gait, M1 including four plantar flexors (soleus, gastrocnemius medialis, gastrocnemius lateralis, peroneus longus) and M3 including two hamstrings (biceps femoris, semitendinosus). For gait stop and balance, in addition to M1 and M3, a third muscle synergy, M2, including three quadriceps (rectus femoris, vastus medialis, vastus lateralis) could be distinguished. For sit to stand and stand to sit it was possible to observe M2 and M3.

Despite significant limitations, it is possible to conclude that muscles from the same group showed similar patterns of activation in healthy subjects, emphasizing the importance of per- forming further investigations.

(42)

BIBLIOGRAPHY

Bibliography

[1] Allan Colver, Charles Fairhurst, and Peter O D Pharoah. “Cerebral palsy”. In: The Lancet 383.9924 (2014), pp. 1240–1249.

[2] Neil Wimalasundera and Valerie L Stevenson. “Cerebral palsy”. In: Practical Neurology 16.3 (2016), pp. 184–194.

[3] Roland Thietje and Sven Hirschfeld. “Epidemiology of Spinal Cord Injury”. In: Neuro- logical Aspects of Spinal Cord Injury. Ed. by N. Weidner and al. Springer International Publishing Switzerland, 2017. Chap. 1, pp. 3–17.

[4] Yannick Béjot, Henri Bailly, Jérôme Durier, and Maurice Giroud. “Epidemiology of stroke in Europe and trends for the 21st century”. In: La Presse Médicale 45.12, Part 2 (2016).

QMR Stroke, e391–e398.

[5] A. M. Dollar and H. Herr. “Lower Extremity Exoskeletons and Active Orthoses: Chal- lenges and State-of-the-Art”. In: IEEE Transactions on Robotics 24.1 (Feb. 2008), pp. 144–

158.

[6] Zachary F. Lerner, Diane L. Damiano, and Thomas C. Bulea. “The Effects of Exoskeleton Assisted Knee Extension on Lower-Extremity Gait Kinematics, Kinetics, and Muscle Activity in Children with Cerebral Palsy”. In: Scientific Reports 7.1 (Oct. 2017).

[7] Steven E. Irby, Kathie A. Bernhardt, and Kenton R. Kaufman. “Gait of stance control orthosis users: The Dynamic Knee Brace System”. In: Prosthetics and Orthotics Interna- tional 29.3 (Dec. 2005), pp. 269–282.

[8] Max K. Shepherd and Elliott J. Rouse. “Design and characterization of a torque-controllable actuator for knee assistance during sit-to-stand”. In: 2016 38th Annual International Con- ference of the IEEE Engineering in Medicine and Biology Society (EMBC). IEEE, Aug.

2016.

[9] Max K. Shepherd and Elliott J. Rouse. “Design and Validation of a Torque-Controllable Knee Exoskeleton for Sit-to-Stand Assistance”. In: IEEE/ASME Transactions on Mecha- tronics 22.4 (Aug. 2017), pp. 1695–1704.

[10] Yuri P. Ivanenko, Germana Cappellini, Nadia Dominici, Richard E. Poppele, and Francesco Lacquaniti. “Coordination of Locomotion with Voluntary Movements in Humans”. In:

Journal of Neuroscience 25.31 (2005), pp. 7238–7253.

[11] Gelsy Torres-Oviedo and Lena H. Ting. “Muscle Synergies Characterizing Human Postu- ral Responses”. In: Journal of Neurophysiology 98.4 (2007). PMID: 17652413, pp. 2144–

2156.

[12] Lena H Ting and J Lucas McKay. “Neuromechanics of muscle synergies for posture and movement”. In: Current Opinion in Neurobiology 17.6 (2007). Motor systems / Neurobi- ology of behaviour, pp. 622–628.

[13] David J Clark, Lena Ting, Felix E Zajac, Rick Neptune, and Steven Kautz. “Merging of Healthy Motor Modules Predicts Reduced Locomotor Performance and Muscle Coordi- nation Complexity Post-Stroke”. In: 103 (Dec. 2009), pp. 844–57.

[14] Seyed A Safavynia, Gelsy Torres-Oviedo, and Lena Ting. “Muscle Synergies: Implications for Clinical Evaluation and Rehabilitation of Movement”. In: 17 (May 2011), pp. 16–24.

[15] Vicon Motion System. Modeling with Plug-in Gait. Online; last checked, 21-March-2018.

2018. url: https : / / docs . vicon . com / display / Nexus25 / Modeling + with + Plug - in+Gait.

[16] Nasrul Abd Razak, Noor Abu Osman, Hossein Gholizadeh, and Sadeeq Ali. “Development and performance of a new prosthesis system using ultrasonic sensor for wrist movements:

a preliminary study”. In: BioMedical Engineering OnLine 13.1 (2014), p. 49.

[17] Seniam Group. SENIAM Project. Online; last checked, 21-March-2018. url: http : / / www.seniam.org/.

[18] Vicon Motion Systems Ltd UK. Nexus. Online; last checked, 30-April-2018. url: https:

//www.vicon.com/products/software/nexus.

28

(43)

CHAPTER 5. CONCLUSION

[19] Prophysics and Myon. proEMG. Online; last checked, 30-April-2018. 2016. url: http:

//www.prophysics-sol.se/our-products/.

[20] Vicon Motion Systems Ltd UK. Plug-in Gait Dynamic pipeline. Online; last checked, 30- April-2018. 2016. url: https://docs.vicon.com/display/Nexus25/Plug- in+Gait+

Dynamic+pipeline.

[21] Steven E. Irby, Kathie A. Bernhardt, and Kenton R. Kaufman. “Gait of stance control orthosis users: The Dynamic Knee Brace System”. In: Prosthetics and Orthotics Interna- tional 29.3 (2005). PMID: 16466156, pp. 269–282.

[22] Felipe Costa Alvim. c3d2OpenSim. Online; last checked, 22-March-2018. 2014. url: https:

//simtk.org/projects/c3d2opensim.

(44)

(45)

Appendices

(46)

(47)

Appendix A

Literature study

A.1 Introduction

A literature study has been performed in order to establish the current technologies and research trends in neurorehabilitation and motion therapy. The first section (A.2) gives an overview of three medical conditions which affect the motor system. People suffering from those condition are shown to potentially benefit from a newly designed robotic EMG-based knee assistive device. The second section (A.3) shows the most popular assistive devices for lower limb, with a particular focus on knee assistive devices. The third section (A.4) illustrates the concept of muscle synergy, which is the focus of the main thesis project.

A.2 Medical considerations

A.2.1 Spinal cord injuries

The spinal cord is divided into four main segments known as cervical, thoracic, lumbar and sacral, from the brainstem to the lumbar region of the vertebral column. There are seven cervical vertebrae, twelve thoracic vertebrae and five lumbar vertebrae with eight cervical nerves (C1-C8), twelve thoracic nerves (T1-T12) and five lumbar nerves (L1-L5), plus five sacral nerves (S1-S5) branching out from the sacrum and one coccygeal nerve (Co1) exiting the coccyx. The cervical division is responsible for breathing, heart rate and arm movements from the shoulder to the fingers. The thoracic division is responsible for the trunk stability and regulates the sympathetic system. The lumbar division controls the functions of the lower limb, i.e. hip, knee and foot motion, with the addition of S1 for the latter. The sacral nerves control bowel and bladder activities.^1,2

A spinal cord injury (SCI) is a damage to the spinal cord which can affect the functions of the body part controlled by the nerves below the site of injury. For example, a lesion of the cervical segment can result in the complete paralysis (tetraplegia) of the body, while a injury of the lumbar segment will affect only the lower limbs, without altering the functions of the upper body. The lesions can be permanent or temporary and can cause muscle dysfunction and partial or complete loss of sensation. Sensorimotor alterations can lead to serious complications and organ failure (e.g. lungs, bowel, bladder etc.), depending on the gravity and location of the injury.³

Assessing the worldwide incidence of SCIs can be difficult due to the high level of traumatic injuries compared to the non-traumatic or disease-related ones. The statistics are also affected by the social and economical development of the different countries analysed.⁴ Traumatic SCIs consist mostly of road traffic accidents, which are connected to external factors as quality of the roads, traffic control systems and amount of drivers on the roads, which differ from country to country. Falls, violence and sport accidents are other causes of traumatic SCIs. The prevalence of traumatic SCIs plays an important role in the treatment of the lesions, since traumas affect people of all ages and genders, often causing a permanent life change. It is estimated that 0.1%

of the world population is affected by disabilities due to SCIs. This value must be considered carefully due to the lack of a global register and the fact that these numbers are retrieved from

(48)

A.2. MEDICAL CONSIDERATIONS APPENDIX A. LITERATURE STUDY

valid data of very few countries (USA, Canada, Norway, Finland, Canada, Australia, Germany, France).⁴ Non-traumatic SCIs are caused by diseases as infections and tumors which mainly affect elderlies.⁴

Current neurorehabilitation strategies

For patients with incomplete SCIs, i.e. when the brain still has the ability to send signals to the part below the site of the lesion, the therapy is directed towards functional independence.

The main goal is for the patient to walk again, which can be achieved through task-specific rehabilitation, i.e. ”if you want to walk, you have to walk”.⁵ This has to be combined with training of the upper body to sustain balance and improve control and stability during the gait.

Examples of training strategies include treadmill training, with or without the use of body- weight support devices, depending on the level of the injury. Overground training naturally follows the treadmill therapy or can be carried out as a first treatment, if the condition of the patient allows it.

Body weight supported training can be either manual or robotic-assisted.⁵ The first requires for the therapist to be strong enough to support the patient, increasing the risk for further injuries for both the patient and the therapist. For this reason and in order to provide a more efficient therapy, robotic body weight supported training has become more popular in rehabilitation treatments. Examples of this technology are ERIGO^®, LOPES^®, ALEX^®, Lokomat^®(Fig. A.1), G-EO-System, LokoHelp, Haptic Walker, Gait Trainer GT1 and More- Gait. These are all treadmill-based training therapies so they forego, in general, the overground training, which will eventually lead to independent motion. Overground training can also be robotic-assisted, by systems like the LiteGait and the Zero G.⁵ The final goal is independent walking, when possible, or wearable-device-assisted walking.

Figure A.1: Lokomat Pro (Hocoma AG, Volketswil, Switzerland).⁶ An example of treadmill training.

34

(49)

A.2.2 Cerebral palsy

Cerebral palsy (CP) identifies a group of disorders of movement and posture.^7,8 It affects 0.2- 0.3% of infants and children of age <2 in Europe and the US and it is a life-long disease since there is no cure.^7,8 CP is caused by brain injury during pregnancy, at the time of delivery or during the first years of a child’s life (up to the age of 2⁷). CP does not only cause motor impairment but also ”disturbances of sensation, perception, cognition, communication, behaviour, by epilepsy and by secondary musculoskeletal problems.”⁹The classification of the CP syndrome varies according to the location of the injury in the brain (cerebral cortex, cerebellum etc.), its symptoms (spasticity, ataxia etc.), the distribution of limb impairment (hemiplegia, quadriplegia etc.), the degree of muscle tone (iso-, hypo-, hyper- tonic) and the time of the injury (prepartum, delivery, etc.). This classification is extremely important during the diagnosis since it defines interventions and treatments for the patients. Life expectancy and quality of life depend on the gravity of the motor and cognitive impairments and are strictly related to the quality of the treatments that the patients undergo throughout their entire life. The sever- ity of the motor impairment is established by the Gross Motor Function Classification System (GMFCS) (Fig. A.2) and the Manual Ability Classification System (MACS).

Crouch gait and posture

One of the most common gait and posture pathologies is crouch gait or crouched posture. It consists on emphasized knee flexion at the stance phase and consequent stressed hip flexion and hip rotation. This pathology worsens over time and can be the cause of further joint degeneration, pain, muscle weakness, spasticity, musculoskeletal deformities and motor control deficits.¹⁰Several studies contributed to outline the muscle adaptation patterns underlying the crouch gait.¹¹Compared to normal gait, crouch gait requires higher knee extensor strength and lower strength for ankle plantar flexors and hip abductors. This suggests that crouched posture and gait are the result of muscle weakness of plantaflexors and abductors, weakness that is worsened with age when maintaining the crouched posture.¹²

(50)

GMFCS E & R between 6

^th

and 12

^th

birthday:

Descriptors and illustrations

GMFCS Level V

Children are transported in a manual wheelchair in all settings. Children are limited in their ability to maintain antigravity head and trunk postures and control leg and arm movements.

GMFCS Level IV

Children use methods of mobility that require physical assistance or powered mobility in most settings. They may walk for short distances at home with physical assistance or use powered mobility or a body support walker when positioned. At school, outdoors and in the community children are transported in a manual wheelchair or use powered mobility.

GMFCS Level III

Children walk using a hand-held mobility device in most indoor settings. They may climb stairs holding onto a railing with supervision or assistance. Children use wheeled mobility when traveling long distances and may self-propel for shorter distances.

GMFCS Level II

Children walk in most settings and climb stairs holding onto a railing. They may experience difficulty walking long distances and balancing on uneven terrain, inclines, in crowded areas or confined spaces.

Children may walk with physical assistance, a hand- held mobility device or used wheeled mobility over long distances. Children have only minimal ability to perform gross motor skills such as running and jumping.

GMFCS Level I

Children walk at home, school, outdoors and in the community. They can climb stairs without the use of a railing. Children perform gross motor skills such as running and jumping, but speed, balance and coordination are limited.

GMFCS descriptors: Palisano et al. (1997) Dev Med Child Neurol 39:214–23

CanChild: www.canchild.ca Illustrations Version 2 © Bill Reid, Kate Willoughby, Adrienne Harvey and Kerr Graham, The Royal Children’s Hospital Melbourne ERC151050

Figure A.2: Gross Motor Function Classification System Expanded & Revised (GMFCS E&R) for children of age 6-12.¹³A similar scheme is used for children of age 12-18.

36

(51)

A.3. LOWER LIMB ASSISTIVE DEVICES APPENDIX A. LITERATURE STUDY

A.2.3 Stroke

A stroke is identified as a medical condition in which damage or death of brain cells is caused by either rupture of blood vessels and consequent brain bleeding or by a blockage of the blood flow to a part of the brain.¹⁴ Studies on the epidemiology of stroke show that stroke is the main cause of adult disability in the U.S., and the fourth cause of death with a prevalence of 3% of the population.¹⁵ Studies conducted in Europe show a prevalence of 1.34% strokes every year.¹⁶ The worldwide incidence of stroke depends on several risk factors such as age, gender, race ethnicity, geography and heredity.¹⁵ Other contributing factors are health-related, such as hypertension, cardiovascular diseases, smoking, diabetes and alcohol consumption. The brain damage caused by a stroke can result in muscle weaknesses, paralysis, aphasia, swallowing impairment, memory loss and bladder-related problems.¹⁷ All these symptoms depends on the location and gravity of the stroke. In many cases, about 50-70% of the patients, people regain their functional independence. However, the recovery is often incomplete and most of the patients have life-changing disabilities.

Recovery after stroke

The rehabilitation strategies generally follow a phase-dependent scheme. The four phases of a stroke are divided in hyperacute (<24 hours after stroke), acute (24 hours to 4 weeks), subacute (1 to 3 months) and chronic (>3 months).¹⁷

Different clinical trials have shown the importance of early mobilization during the hyperacute phase, demonstrating that the recovery of motor skills is faster and more complete if mobility treatments start almost immediately after the stroke. Patients surviving a stroke, usually suffer from hemiplegia, i.e. the paralysis of one side of the body. Both upper and lower extremities of either the left or right side are affected and must be treated simultaneously.

Motion therapy must be continuous throughout the four phases and it is aimed at functional independence.

During the acute phase, physical therapy is accompanied by psychological treatment to help patients suffering from post-stroke depression.

The subacute phase focuses on the improvement of the motor skills and the reduction of disability. The urge to restore motor functions is due to the capability of the human body to adapt, e.g. to overcome the lack of functions of one impaired limb the human body is able to find new adaptation strategies and perform a specific motion in a new, different way. This can lead to unconventional movements and the recruitment of different muscle units and might result, in the long term, to further injuries. For example, during walking the flexion and extension of the knee is essential for a normal gait. However, when knee flexion/extension is prevented, it is possible to perform a step by increasing hip adduction/abduction. In the long term, this walking pattern can lead to hip, pelvis and back problems. Studies have shown the benefits of the constraint-induced therapy (CIT) in the functional recovery of motor skills during the subacute phase. This strategy consists in constraining the nonparetic limb (the ”healthy” one) to force the patient to use the impaired one and restore its function by stimulating the recovery of the damaged brain tissue.

During the chronic phase, muscle spasticity can appear. Spasticity can be identified as the uncontrolled, increased muscle activity, caused by an imbalance in the electric signals between the nervous system and the muscles.¹⁸Muscle spasticity is mostly treated with oral medication as nerve blocks.

As already stated, not many stroke survivors regain their complete motor functions to a full extent as before the stroke. This results in permanent disabilities or extremely long rehabilitation therapies.

A.3 Lower limb assistive devices

Assistive devices can be classified into orthoses, exoskeletons and prostheses. An orthosis is an external device aimed to restore the natural function of an impaired limb.¹⁹ Historically, the term exoskeleton has been used to define external devices aimed to augment the performance of healthy subjects in motion related activities. However, due to the great potential that these device have shown in rehabilitation therapies, the term exoskeleton is now used in parallel to

Lower Limb Muscle Synergy During Daily Life Activities: A Way to Convey Intended Motions To a Robotic Assistive Device.

Lower limb muscle synergy during daily life activities.

A way to convey intended motions to a robotic assistive device.

TERESA COLANGELO

Lower limb muscle synergy during daily life activities.

A way to convey intended motions to a robotic assistive device.

TERESA COLANGELO

Abstract

Acknowledgements

Contents

List of Figures

List of Tables

List of Abbreviations

Chapter 1

Introduction

1.1 Aim

1.2 Hypothesis

Chapter 2

Methods

2.1 Experimental setup and protocol

2.2 Data processing and analysis

Chapter 3

Results

3.1 Gait

3.2 Gait-stop and balance

3.3 Sit-to-stand

3.4 Stand-to-sit

Chapter 4

Discussion

4.1 Main Findings

4.1.1 Gait

4.1.2 Gait-stop and balance

4.1.3 Sit-to-stand

4.1.4 Stand-to-sit

4.2 Limitations

4.3 Future work and possible applications

Chapter 5

Conclusion

Bibliography

Appendices

Appendix A

Literature study

A.1 Introduction

A.2 Medical considerations

A.2.1 Spinal cord injuries

A.2.2 Cerebral palsy

GMFCS E & R between 6

and 12

birthday:

Descriptors and illustrations

A.2.3 Stroke

A.3 Lower limb assistive devices