Quanti cation of motor patterns in parkinsonian gait
Transcript of Quanti cation of motor patterns in parkinsonian gait
Quantification of motor patterns inparkinsonian gait
Luisa Fernanda Cardenas Martınez
Cim@Lab Research GroupUniversidad Nacional de Colombia
School of Medicine
Bogota, Colombia
2016
Quantification of motor patterns inparkinsonian gait
Luisa Fernanda Cardenas Martınez
Thesis presented as partial requirement for the degree of:
Master in Biomedical Engineering
Advisor:
Eduardo Romero Castro. MD. Ph.D.
Research Area:
Human Gait Analysis, Movement Analysis
Cim@Lab Research Group
Universidad Nacional de Colombia
School of Medicine
Bogota, Colombia
2016
“Sabemos lo que somos, pero no lo que podemos ser”
William Shakespeare
Acknowledgments
To my family, for their unconditional support, love and motivation to keep going.
To the professor Eduardo, for giving me the chance to do something different, and for pushing
me to think in a different way
To Fabio Martinez, for taking the challenge of teaching engineering concepts to physios. for
the great patience and his friendship.
To “el flaco” and “Truji”, for being the best psychologists, comedians,....in general the best
friends .
And to Cim@lab, for every day in the lab, that allows me to learn .
vi
Abstract
Worldwide Parkinson’s disease (PD) is the second most common neurodegenerative dis-
ease with around 6.3 million cases diagnosed. This disease is characterized by a progressive
disorder of the nervous system, mainly manifested in gait and postural control impairment
that alter the functional independence in patients. Gait analysis becomes as an important
tool to determine the severity of disease as well as to orient the diagnosis. Nevertheless, such
analysis is currently highly dependent on the physician expertise as to detect some subtle
characteristic movements of the disease. Physical gait models arise as a powerful alternative
to reduce the variability in movement analysis. These models allow to emulate the progres-
sion and performance of gait patterns in different disease stages, providing information for
the selection of appropriate treatments. This thesis presents a physical model that integrates
the estimation of the center of gravity (CoG) trajectory based on trunk kinematics, and the
quantification of the symptoms severity for different PD stages. A physical gait model is
herein emulated by using a double inverted pendulum for the leg swinging and a damper-
spring System (SDP) to recreates the double support phase. This model is then adjusted
to follow a particular CoG learned for each stage. The evaluation of the CoG estimations
performed by trunk patterns was performed by comparing the resulting trajectories with
actual ones learned from subjects in stages 2, 3, and 4, achieving a correlation coefficient
of 0.88, 0.92 and 0.86, respectively. The parameters from the model were adjusted to gen-
erate similar trajectories to the ones previously estimated. The frechet distance between
the trajectories generated by the model and the CoG estimation is 0.07, 0.09, 0.29 for each
stage.
Keywords: Parkinson’s disease, Pathologic gait patterns, Gait modelling, trunk
bending patterns.
vii
Resumen
A nivel mundial, la enfermedad de parkinson (EP) es la segunda enfermedad degenerativa
mas comun, con cerca de 6.3 millones de casos diagnosticados. Esta enfermedad se carac-
teriza por un desorden progresivo del sistema nervioso que se manifiesta principalmente en
alteraciones de la marcha y del control postural que limita la independencia funcional de
los pacientes. El analisis de los patrones en la marcha se ha convertido en una herramienta
util para determinar la severidad de la enfermedad, ası como para orientar el diagnostico.
Sin embargo, este analisis es altamente dependiente de la experticia del evaluador, especial-
mente para la deteccion de algunos movimientos sutiles caracterısticos de la enfermedad. Los
modelos fısicos emergen como una alternativa sobresaliente para disminuir la variabilidad
en el analisis de movimiento. Los modelos permiten emular la progresion y los patrones de
la marcha en diferentes momentos de las enfermedades, proporcionando informacion para
la seleccion de tratamientos adecuados. Esta tesis presenta un modelo fısico que integra
la estimacion de la trayectoria del centro de gravedad (CoG) basada en la cinematica del
tronco y la cuantificacion de la severidad de los sıntomas en los diferentes estadıos de la EP.
El modelo fısico propuesto utiliza un doble pendulo invertido para la fase de apoyo simple
y un sistema resorte-amortiguador para la fase de doble apoyo. Este modelo es ajustado
para seguir un CoG especıfico, definido para cada uno de los estadios. La evaluacion de la
estimacion del CoG realizada por los patrones de tronco se realizo comparando las trayec-
torias resultantes con trayectorias aprendidas de sujetos en estadios 2, 3 y 4. Se obtuvo un
coeficiente de correlacion para la estimacion del CoG de 0.88, 0.92 y 0.86, respectivamente.
Los parametros del modelo fısico se ajustaron para generar trayectorias cercanas a los CoG
estimados anteriormente, obteniendo una distancia de frechet entre las curvas generadas por
el modelo y la trayectoria estimada del CoG de 0.07, 0.09, 0.29.
Palabras claves: Enfermedad de parkinson, Patrones de marcha patologicos,
Modelamiento de la marcha, Inclinacion de tronco.
Contents
Acknowledgments v
Abstract vi
1 Theoretical Framework 2
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Balance assessment in Parkinsonian gait . . . . . . . . . . . . . . . . 3
1.1.2 Quantification and gait modeling . . . . . . . . . . . . . . . . . . . . 3
2 Fusing trunk bending patterns 6
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Proposed Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 The Physical Gait model . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Learning a CoG from trunk bending patterns . . . . . . . . . . . . . 9
2.2.3 Dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Evaluation and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Quantification of gait patterns 14
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 The CoM trajectory and its incidence in the evaluation of PD progression . . 16
3.2.1 The role of Physical Gait models at simulating pathological patterns 16
3.2.2 Damper-Spring Pendulum (SDP) . . . . . . . . . . . . . . . . . . . . 17
3.2.3 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4 Conclusions and Perspectives 22
1 Theoretical Framework
1.1 Introduction
Parkinson’s disease (PD) has been diagnosed in around of 6.3 million people worldwide
[17, 38], representing a major public health challenge. This disease nowadays implies one of
the most high healthcare expenses because of the reduction of productivity, drug treatment,
the outpatient care, among others [50]. PD is characterized by the lost of dopamine neuro-
transmitter that affects the connection of brain areas located in the basal ganglia, producing
inability to control and coordinate movements [42, 58]. In fact, the effect of the decreas-
ing of dopamine concentration in around of 60 − 70 % can be observed as resting tremor,
rigidity, bradykinesia and balance impairment [7, 33]. Hence, the PD diagnosis is based on
the analysis of the sum up of symptoms as the reduction of the frequency and amplitude of
spontaneous movements, as well as, postural instability during positions against gravity.
Classically, the motor changes produced by PD progression have been evaluated by clinical
rating scales as the Hoehn and Yahr staging [3]. This scale defines the motor decline by
setting five broad categories that highlight the characteristics of movement alteration with
the major impact in the performance of daily living activities. In the initial stage the motor
impairment appear focally in one limb while for the second stage is defined for bilateral
impairment [61]. The third and fourth stages are characterized by moderate and severe
affectation of balance. In the advanced five stage the motor impairments alter the gait
performance and it is necessary a external aid. The assessment of PD progression with this
scale allows to register only coarse motor changes [70, 43], limiting the analysis of motor
patterns among the stages and introducing a natural inter and intra observer variability.
Consequently, Gait analysis has been an important tool in the assessment of motor patterns
in PD progression. In such case, a set of locomotion features such as, swing time, stride
time, and left/right asymmetry, are coded to describe a particular gait PD pattern [4].
This evaluation is commonly carried out by observing the locomotion of a patient during
a short period of time and namely under very restrictive scenarios, i.e., reduced space.
These conditions limit the detection of subtle motion changes, specially in a large number of
patterns[52]. One of main drawback in human gait analysis is that only consider local and
independent patterns of the lower limb dynamics. Additionally, such analysis ignores the
interaction with the other parts of the body as axial structures (trunk) which can be useful
to determine balance and equilibrium status or other associated impairments.
1.1 Introduction 3
1.1.1 Balance assessment in Parkinsonian gait
The CoG response during time is associated to the alternated pendular movement of the
lower limbs, being the main pattern studied in gait analysis. Despite of the importance of
lower limbs to produce motion, current analysis omit the force interaction produced by the
central and upper part of the body. In fact, under the control loop to maintain a stable and
efficient gait, the upper body play a fundamental role because the trunk muscles minimize
the movement produced by the lower limbs, stabilizing the spine and the head [9]. Some
authors have shown that this trunk activation anticipates imbalance due to gait performance
and stabilizes spine segments in a top/down manner(from cervical to lumbar spine)[54].
The balance impairment, defined mainly as the losing of muscles strength in trunk, alters
the ability to maintain the body’s equilibrium under static and dynamic conditions. This
alteration constitutes the cardinal motor symptom in PD, namely present in populations with
the faster evolution and lower response to the pharmacological treatment [19]. This balance
lost is progressive altered and increase the risk of falling [51], and reduce the quality of life
[2, 62]. Trunk bending observations such as forward trunk inclination, lateral bending and the
pelvic obliquity are essential patterns that allows to characterize balance state and coarsely
determine intervention strategies [26, 65],[8],[66]. Currently, such patterns are measured in
sagittal and coronal plane under specific tasks like sitting or standing to assess postural
instability in PD [1]. However, these preliminary studies lost important correlation of such
patterns with lower limbs variables during gait, for instance, the dynamic stabilization cost
by abnormal behaviors of the trunk [68].
1.1.2 Quantification and gait modeling
The relationship between trunk bending patterns and lower limbs kinematics allows for
instance the understanding of the coordinated actions between the stabilization produced in
axial body segments and the dynamics of distal body parts. To describe these axial and distal
patterns during gait, there is available diverse kind of information such as kinematic, that
register the body segments motion in space, kinetic, that quantify the relations between
action-reaction forces, moments and powers of each body segment and muscle activation
variables that evaluate the electrical muscle activity[11]. These evaluations produce several
data from different structures that have to be compared and correlated to obtain useful
information to take actions in clinical practice. However, as it is a demanding task, it is not
usually consider to plan treatments or register results. Then, the assessment of gait patterns
falls in the expertise of the physician when observing patient’s walk. Human gait models
emerge as an option to obtain reproducible and objective information that might predict
and emulate the patient patterns, supporting thereby diagnosis and gathering together the
elements to properly follow up the disease evolution.
Many gait models with different complexity levels have been introduced in the literature,
4 1 Theoretical Framework
some of the more complex are neuromuscular models that are based on the concept of a
central pattern generator(CPG) responsible for the generation and control movement. This
concept motivate the emulation of human neural circuits that can produce rhythmic pat-
terns of neural activity without receiving rhythmic inputs[32]. Although some applications
in dynamic simulators uses this concept for biomedical problems, the concept of the CPG
in human locomotion is a matter of discussion. On the other hand, Musculoskeletal com-
puter models consider the musculoskeletal system as complex mesh of muscles and bone
segments, and take into account physical aspects such as muscle tension and parameters
that represent muscle properties (length, strength), obtained experimentally from dynamic
electromyography and the measured torque[14, 31]. These models require many real data
and expert knowledge to tune manually a large number of parameters, limiting their use in
routine clinical practice.
The most basic gait models are dedicated to describe the global body dynamics by using
physical representations such as a pendulum model or a mass-spring systems [34]. The
pendulum-gait model built from stiff-legged [40] simulates the center of mass (CoM) trajec-
tory during a period of the stance limb, allowing to establish a basic energy transfer analysis
during motion. This model is a very basic representation which misses the leg description and
only involves a structural rigid representation. A complementary mechanism that represents
the lower limb structure as a coupled pendulum has been also proposed [21], but under the
strong assumption that the human locomotion only depends on some basic mass and length
relationships. Alternatively, motion models based on spring-mass systems have been also
used for representing the gait but only when particular patterns are described, such as the
heel strike and the double support phase [40, 25]. Geyer et al [24, 37] proposed a structural
gait model including a pendular system, coupled to a spring mass and a radial structure that
reduces the friction during movement, allowing to analyze how the speed variation during
normal gait affects the leg stiffness. This simple model properly fitted the walking in young
and old people, observing that stiffness and damping may increase with the speed, but with
greater variations in older people [30].
This thesis has contributed to the construction of a computational model to quantify motor
patterns in parkinsonian gait, using complementary information from the upper part of the
body. It provides information that allows the analysis, interpretation and visualization of gait
disturbances in the progression on PD. As a first step, the strategy consists in introducing
trunk bending information to changes in the CoG trajectory, see chapter 2. A contribution
weight was obtained from a linear combination of three trunk patterns for the estimation of
the CoG in each PD stage. Afterwards, this CoG trajectory is tracked by a physical gait
model in the whole gait cycle, where, the model parameters (k and b) are used to aid the
clinician to quantitatively describe this pathology, see chapter 3. The physical gait model
consist in coupling two different systems, the double inverted pendulum that emulates the
leg swinging for the single support phase and a damper-spring System (SDP) that recreates
1.1 Introduction 5
both legs in contact with the ground for the double phase. The model parameters obtained
show discriminating results among the stages and have a physiological representation.
2 Simulation of parkinsonian gait by
fusing trunk learned patterns and a
lower limb first order model
Presented on the “10th International Symposium on Medical Information Processing and
Analysis” SIPAIM 2014, October 2014
Parkinson’s disease is a neurodegenerative disorder that progressively affects the movement.
Gait analysis is therefore crucial to determine a disease degree as well as to orient the
diagnosis. However, gait examination is completely subjective and therefore prone to errors
or misinterpretations, even with a great expertise. In addition, the conventional evaluation
follows up general gait variables, which amounts to ignore subtle changes that definitely
can modify the history of the treatment. This work presents a functional gait model that
simulates the center of gravity trajectory (CoG) for different parkinson disease stages. This
model mimics the gait trajectory by coupling two models: a double pendulum (single stance
phase) and a spring-mass model (double stance). Realistic simulations for different parkinson
disease stages are then obtained by integrating to the model a set of trunk bending patterns,
learned from real patients. The proposed model was compared with the CoG of real parkinson
gaits in stages 2, 3, 4 achieving a correlation coefficient of 0.88, 0.92 and 0.86, respectively.
2.1 Introduction 7
2.1 Introduction
Parkinson’s disease (PD) is a neurodegenerative disorder that affects more than 6.3 million
of people worldwide[29]. A main clinical manifestation of this disease is the alteration of
gait patterns, namely: shuffling steps, reduced walking speed and loss of balance. This last
alteration is produced by the poor response of certain muscles and the deficient coordination
between the structural system and neuromotor commands, leading to an alteration of the
posture when the trunk bends forward. Normal and pathological gaits include complex inter-
actions between the neuromuscular, the musculo-tendinous and the osteoarticular systems
[18],[27]. In spite of this complexity, the following up and prognosis of parkinson disease
are limited by the physician experience, introducing the consequent inherent inter-expert
variability [69]. Hence, reliable, efficient and reproducible physical gait models are required
to quantify the disease and improve the treatment planning of parkinson disease.
Many models have been proposed in diverse areas[47],[13] to describe normal gait patterns.
In general physical representation are used to represent the dynamic of musculoskeletal
structure and a coupling control that emulate the neuromotor commands [59],[63]. However
, these models are limited to simple and restrictive representations of normal trajectories
that do not include abnormal or pathological patterns. An usual gait model is the coupled
pendulum system which mimics the single stance phase but misses important gait phases such
as the heel strike and the double stance phase [21]. The mass-spring system has simulated
human movements such as running, but under some restrictions w.r.t to real gait patterns
[55]. Complementary approaches [48] couple pendulum and physical models to represent
normal and pathological movements. Nevertheless this gait model learns only from lower
limb patterns, ignoring important motion features of the parkinson disease like the trunk
bending.
The main contribution of this work is a coupled physical gait model that learns trunk bend-
ing patterns and simulates different stages of the parkinson disease. The proposed model
mimics the single stance phase with an oscillating trajectory represented as a coupled dou-
ble pendulum while the double stance phase is approximated by a spring mass system. The
parameters of the double stance phase are learned from a set of trunk bending patterns.
2.2 Proposed Approach
The pipeline of the proposed approach is illustrated in Figure 2-1. The single stance is
represented by a coupled-pendulum model while the double stance is simulated as a spring
mass system. Then, the parameters of the model are learned from a linear combination of
trunk bending patterns.
8 2 Fusing trunk bending patterns
Figure 2-1: Proposal Method Pipeline
2.2.1 The Physical Gait model
The gait Cycle is composed of two phases: The single support, which constitutes the 70-
80% of the cycle, and the double support that constitutes about 20-30%. In this work, a
lower limb gait model [48] fully describes the gait cycle during the single support using a
coupled pendulum trajectory and a spring-mass system during the double stance. These
representations are as follows:
The Single Stance Phase: a pendular CoG trajectory
During the single gait stance, a fixed limb supports the body while the other limb swings
forward, using the hip as a fix pivot . A coupled double pendulum system simulates such
movement, See figure 2-1(a), i.e., the mass m, representing the lower limbs, moves forward
a mass M , the upper part, producing the illustrated oscillatory CoG trajectory. The gait
model corresponds to a coupled system of two non-linear differential equations:
β(1 − cosφ)(3θ − φ) − β sinφ(φ2 − 2θφ) + (g sin θ
l)(β(sin(θ − φ) − 1)) = 0
θ(β(1 − cosφ)) − βφ+ βθ2 sinφ+ (βg
l) sin(θ − φ) = 0 (2-1)
where β = m/M , θ is the angle of the stance leg at the particular time t with respect to the
slope and φ is the angle between the stance leg, l is the leg length and g is gravity.
2.2 Proposed Approach 9
The Double stance phase
During the double stance, a forward momentum of the body is produced when the limb
contacts the ground. In this phase, the hip is initially flexed, facilitating the propulsion of
the (opposite) fixed limb (toe-off). The equation is:
Mx = llx− lr(d− x)
My = lry + lry − gM (2-2)
where g is the gravity,d is the inter-leg distance, x and y are the positions in x and y planes,
ll and lr are the left and right legs, respectively and their length changes are:
ll = k(l0√
x2 + y2 − 1)
lr = k(l0√
(d− x)2 + y2 − 1) (2-3)
where k is the spring constant that determines the change of the leg length during the double
support, modeling certain parkinson patterns, for instance, the speed and step length.
2.2.2 Learning a CoG from trunk bending patterns
The balance impairment, a main PD sign, is produced by a progressive weakness of some
postural muscles, debilitating postural adjustments [60] . In consequence, the trunk bending
results a compensation mechanism to maintain the equilibrium and an important sign to
characterize parkinson disease stages [46, 44, 36]. By calculating the trunk bending, it is
possible to follow the progressive muscle imbalance that generates a lateral deviation and
furthermore, to describe the imbalance of the lower limbs by the pelvis obliquity. The
herein proposed work achieves much more accurate PD simulations by introducing trunk
bending information. For doing so, the trunk forward inclination (TFI) x1,t, the trunk
lateral bending(TLB) x2,t, the pelvis obliquity (PO) x3,t were captured and delineated from
the sagittal and coronal views as suggested in [20, 41].
Then, the CoG trajectory was estimated as a linear combination of the three trunk trajec-
tories: CoG(t) =∑
i=1...3Wi,txi,t, where Wi,t is a contribution weight. The estimated CoGCoG(t), for the different PD stages, is then used as a ground truth to map the different CoG
trajectories described by the physical model. A newton-raphson method was used to find the
k parameter describing the most similar CoG w.r.t the learned CoG(t) for the different PD
stages. The similarity between the CoG curves was computed as a simple euclidean metric
of both trajectories, during a gait cycle.
10 2 Fusing trunk bending patterns
2.2.3 Dataset
Data used in this work include a total of 24 patients diagnosed with PD in stages 2, 3
and 4 [53]. These patients were captured in the gait laboratory of Universidad Nacional de
Colombia, under a plug in gait Vicon protocol. For each patient, it was recorded a gait cycle
from IEEE cameras (320 × 240 and 30 fps) in sagittal and coronal planes. Patients signed
an informed consent previously approved by ethics committee of the school of medicine of
Universidad Nacional de Colombia according to the Declaration of Helsinki.
StagesNumber of
AgeVideos
PD 2 6 64.6 ± 7.44
PD 3 7 65.5 ± 5.5
PD 4 4 70.2 ± 8.92
Table 2-1: Dataset characteristics
2.3 Evaluation and Results
The trunk learned parameters by PD stages, are shown in figure 2-2. Evaluation was carried
out by comparing the estimated CoG trajectory w.r.t the average of real CoG trajectories in
the different stages. The model uses the trunk patterns to match a specific CoG trajectory.
The learned k parameters for the model are reported in table 2-2. This k parameter increases
for advanced levels of PD, following the stretch of the limbs and emulating the stiffness and
slow motion of the PD.
PD Stages PD 2 PD 3 PD4
K value estimated 136.2 ± 60.6 197.6 ± 65.1 280 ± 28.28
Table 2-2: Values of constant k estimated by stages
Quantitative evaluation was performed using a standard correlation coefficient between the
estimated CoG and the averaged real trajectories, i.e., the ground truth. The correlation
coefficient measures the trend of two signals and globally estimates their similarity. In Figure
2-3, it is illustrated in blue, the estimated CoG, and in black, the ground truth during a
gait cycle for different PD stages.
2.4 Conclusions 11
Figure 2-2: Trunk learned Parameters by stages
In table 2-3, it was reported the PD correlation for the different stages, showing in average
an 0.891 with a standard deviation of 0.0293. The results obtained in this preliminary work
suggest the model is capable of simulating complex trajectories that may characterize the
disease with only two parameters.
PD Stages PD 2 PD 3 PD 4
Coefficient Correlation 0.8839 0.9242 0.8672
Table 2-3: Correlation coefficient of the COG estimation from trunk data vs real data
2.4 Conclusions
This paper presents a novel approach that simulates the center of gravity trajectory (CoG)
for different parkinson disease stages. The whole strategy consists in estimating the CoG
12 2 Fusing trunk bending patterns
Figure 2-3: Comparison of CoG trajectories between the mean curve estimated from trunk
data and real trajectories obtained from videos
2.4 Conclusions 13
by integrating a set of trunk bending patterns learned from real patients with a very simple
lower limb model. The gait results characterized with a minimal number of parameters that
are entailed with physical meaning and can be used to interpret and correlate different stages
of this complex disease.
3 Quantifying the parkinson’s disease
progression by simulating gait patterns
Presented on the “11th International Symposium on Medical Information Processing and
Analysis” SIPAIM 2015, November 2015
Modern rehabilitation protocols of most neurodegenerative diseases, in particular the Parkin-
son Disease, rely on a clinical analysis of gait patterns. Currently, such analysis is highly
dependent on both the examiner expertise and the type of evaluation. Development of eval-
uation methods with objective measures is then crucial. Physical models arise as a powerful
alternative to quantify movement patterns and to emulate the progression and performance of
specific treatments. This work introduces a novel quantification of the Parkinson disease pro-
gression using a physical model that accurately represents the main gait biomarker, the body
Center of Gravity (CoG). The model tracks the whole gait cycle by a coupled double inverted
pendulum that emulates the leg swinging for the single support phase and by a damper-spring
System (SDP) that recreates both legs in contact with the ground for the double phase. The
patterns generated by the proposed model are compared with actual ones learned from 24 sub-
jects in stages 2,3, and 4. The evaluation performed demonstrates a better performance of
the proposed model when compared with a baseline model(SP) composed of a coupled double
pendulum and a mass-spring system[48]. The Frechet distance measured differences between
model estimations and real trajectories, showing for stages 2, 3 and 4 distances of 0.137,
0.155, 0.38 for the baseline and 0.07, 0.09, 0.29 for the proposed method.
3.1 Introduction 15
3.1 Introduction
Parkinson disease (PD) affects around 6.3 million people worldwide[29]. This pathology is
mainly characterized by movement alterations such as loss of gait automaticity and postural
imbalance [64]. The analysis of the Parkinson gait pursuits the disease characterization and
the follow up of the disease progression[57]. However, this evaluation is highly dependent
on the physician expertise, with the consequent inter and intra observer variability. Fur-
thermore, in practice, the time devoted to clinical examination usually results too short as
to detect subtle motion changes that might be determinant with respect to the disease pro-
gression [52]. Standardization of a particular evaluation requires the development of metrics
that allow an accurate and reproducible comparison between different gaits, a difficult task
when the correspondence between several patterns must be temporal.
Physical gait models emerge as reliable and reproducible methods that can be used as metrics
since they achieve an accurate description of a set of kinematic patterns during locomotion.
The physical models facilitate visualization of biomarkers like the center of gravity, an im-
portant parameter hardly estimated when physicians perform an observational evaluation.
In the literature, many gait models with different complexity levels have been introduced.
Some complex models predict muscle-tendon forces as well as their contribution to joint
accelerations [28]. A main drawback with these models, is that they require an exhaustive
tuning of a large number of parameters as well as prior expert Knowledge e.g. three dimen-
sional bone surface geometry and muscle geometry [45]. This process is highly dependent on
the expert knowledge, while the obtained information is specific for pathology and motion
patterns. On the other hand, low complexity models represent the gait dynamics with global
relationships between very simplified body structures [34] as the inverted pendulum system
[21], still limited when emulating only the simple gait stance. Attempting some improve-
ment for describing the gait, previous works have coupled soft tissue representations to the
inverted pendulum, extending the emulation to the double support phase. These models
have been useful when a general gait description is required, e.g. Geyer et al [24, 37] pro-
posed a pendulum system coupled to a spring mass and a radial foot structure that simulates
the speed variation during normal gait. Afterwards, Hong et al [30] observed the spring-like
leg dynamics in two different age groups, obtaining a different stiffness in both groups. A
spring-mass system coupled to a pendulum, was evaluated in normal and pathological gaits
[48], obtaining different rigidity parameters in normal, PD and crouch gait. These meth-
ods however have been limited to explain normal dynamic patterns, or barely a difference
between normal and pathological stages, but still too far from emulating different PD stages.
This work introduces a novel physical model that determines different PD stages by simu-
lating the gait patterns during a whole gait cycle. The model represents the single stance
phase with a double inverted pendulum while the double stance phase is approximated by
a damper-spring mass system. This paper is organized as follows. Section 2 describes the
proposed method and section 3 presents the results and conclusions.
16 3 Quantification of gait patterns
3.2 The CoM trajectory and its incidence in the
evaluation of PD progression
The CoG is the main gait biomarker, it summarizes main interactions among the body
segments, the external forces and the central neuro-commands[16, 49, 56]. Figure 3-1a
shows the CoG ground truth during a single gait cycle, composed of two steps, one per leg.
Each step starts by the double support phase (DS) followed by the single support(SS) period.
During the DS, the CoG slightly falls down and returns to the initial position, swinging the
body weight from back to front. In the next SS phase, the leg from behind generates a
propulsive force that facilitates the leg elevation and the forward CoG displacement. This
leg then starts the forward swing, that ends when the foot touches the floor.
When compared with normal trajectories, the Parkinson CoG shows a limited vertical dis-
placement for both phases of the cycle. According to Chastan et al[10], this alteration may
be caused by a diminished postural stabilization control and an increased body verticality.
Furthermore, vertical displacement and step duration is asymmetrically observed for both
sides [6]. Hence, the progressive PD motor impairments such as tremor, slowness of move-
ment, stiffness, and balance alterations [67],[39] produce a mark in the CoG trajectory that
become more evident as long as the pathology progresses, see figure 3-1b.
The CoG trajectories in PD were obtained from the analysis of a parkinsonian gait dataset.
An expert tracked the CoG by the sacral marker method[22] and the x, y coordinates were
obtained using an open source video analysis software [12]. The CoG trajectory depends on
the patient height and is expressed as a percentage. The trajectories are normalized in a
range between -1 to 1.
The present work quantifies the PD stages by simulating the CoG using a physical gait
model. For doing so, real CoG trajectories by stages were obtained from a Parkinsonian gait
dataset, and the proposed spring-damper pendulum estimates the CoG trajectories. Finally
the Frechet distance evaluates the similarity between real and the estimated curves.
3.2.1 The role of Physical Gait models at simulating pathological
patterns
Structural physical gait models have emerged as descriptor of the walking dynamic by sim-
ulating the CoG trajectory with known relationships among masses and lengths, describing
motion by simple numerical and deterministic relationships. The main advantage of these
models is that they use few parameters to describe the complex dynamics and establish nat-
ural relationships among multiple displacements and forces. The most basic and important
gait model is the inverted pendulum model[21], a simple mechanical system that represents
the swing of one leg over the other, supporting the entire body weight, representing only the
single support phase. However, subsequent works have used additional systems that comple-
ment the double support phase representation that is missing in the previous approach, e.g.
3.2 The CoM trajectory and its incidence in the evaluation of PD progression 17
(a) Ground truth of normal CoG
trajectory[35].
DS: double support, SS:single
support
(b) Mean curve of the CoG trajectory
in PD
different stages.(PD2,PD3,PD4:
PD stage 2,3 and 4)
Figure 3-1: CoG trajectory in normal and PD
Mass-spring models. These models were initially used to describe hopping and bouncing like
gait models[15],but, Geyer et al [24] in 2006 evidenced that compliant legs in a spring-mass
model, are essential to obtain the basic walking mechanics.
3.2.2 Damper-Spring Pendulum (SDP)
Figure 3-2: Spring-Damper Pendulum model
This model fully describes the center of gravity trajectory during a gait cycle. In the single
support phase, a fixed limb supports the body weight while the other limb swings forward,
using as pivot the mass(M) that represents the hip and the upper part of the body, see figure
3-2. The CoG dynamic in this walking phase is then represented as an oscillatory trajectory
described by a system of two non-linear differential equations as:
β(1 − cosφ)(3θ − φ) − β sinφ(φ2 − 2θφ) + (g sin θ
l)(β(sin(θ − φ) − 1)) = 0
18 3 Quantification of gait patterns
θ(β(1 − cosφ)) − βφ+ βθ2 sinφ+ (βg
l) sin(θ − φ) = 0 (3-1)
where β = m/M , θ is the angle of the stance leg at the particular time t with respect to the
slope and φ is the angle between the stance leg, l is the leg length and g is gravity.
The double stance is represented by a damper-spring mass system. In this phase, the dy-
namic can be understood as the complex interaction of forces produced by passive elements
e.g. tendon, aponeurosis and active elements such as the muscles. From the muscle mod-
elling standpoint, Hill [5] proposed an analogy of these elements with components such as
springs and dampers. The passive elements show elastic characteristics as springs, and with
respect to the active elements, when muscles shorten they spend some of their active force
in overcoming an inherent resistance[23], a similar scenario to what is observed when a pis-
ton shock must be absorbed (damper). In this model, the spring generates the oscillatory
CoG behavior and the damper restrains the excessive CoG motion. The heel strike and the
propulsive action of the opposite limb produces a forward body momentum. The damper-
spring complex represents the leg and receives the body weight, compressed during the first
half of the double stance phase and rebounded during the second half. The equations of
motion for this phase are:
Mx− bx− llx+ lr(d− x) = 0
My − by − lly + lry = Mg (3-2)
where g is the gravity,d is the inter-leg distance, x and y are the positions in x and y planes,
b is the damper constant, ll and lr are the left and right legs, respectively and their length
changes are:
ll = k(l0√
x2 + y2 − 1)
lr = k(l0√
(d− x)2 + y2 − 1) (3-3)
k expresses the spring stiffness, which, determines the change of the leg length during the
double support.
3.2.3 Experimental setup
Data used in this work includes a total of 24 patients diagnosed with PD in stages 2, 3 and 4,
the number of patients by group are 11, 9 and 4 respectively. These patients were captured
in the gait laboratory of Universidad Nacional de Colombia, under a plug in gait Vicon
protocol. For each patient, 6 gait cycles were recorded using IEEE cameras (320 × 240 and
30 fps) in the sagittal plane. Patients signed an informed consent previously approved by
ethics committee of the school of medicine of Universidad Nacional de Colombia, according
to the Helsinki Declaration.
3.3 Results 19
3.3 Results
The evaluation was carried out by the difference of the mean of the CoG estimated from
the model proposed w.r.t the mean of the real curves obtained for each of the PD stages.
Figure 3-3 shows the graphical results per stage, the blue line is the real trajectory and the
red dotted line is the estimation from the model. In stage 2, see figure 3-3a, the estimation
of the model is close to the real trajectory and the slight decrease in the amplitude during
the second part of the gait cycle is properly approximated by the model. In stage 3, see
figure 3-3b this effect is even more evident, the amplitude asymmetry is reproduced by the
model, and also a phase shift in the second half of the gait cycle produced by the earlier
presentation of the CoG ascent during the second cycle half. The real trajectory shows that
one step covers more percentage of the GC than the other. Finally in stage 4, see figure
3-3c, it is observed a larger difference compared with previous stages, however as in PD2,
the SDP model shows a slight amplitude improvement, especially during the double support
phase (0-15 and 50-65 % of GC). In this stage the estimation is a challenge, as the trajectory
does not show a strong periodic and sinusoidal form as for the previous stages.
(a) PD Stage 2 (b) PD Stage 3
(c) PD Stage 4
Figure 3-3: Real CoG trajectories Vs CoG estimation from SP and SDP model for different
PD stages
Quantitative evaluation was performed using the Frechet distance (FD) between the aver-
20 3 Quantification of gait patterns
aged real trajectories, and the estimation of the CoG achieved by the SDP and the baseline
method (SP)[48]. The FD measures the length of the shortest path between two points that
are simultaneously moving through the two curves. The more similar are the two curves,
the smaller the distance and therefore the close to zero they are.
The table 3-1 reported the PD Frechet distance for the different stages. For PD2, the FD in
general shows a difference between the models estimation, having a 0.137 for SP and 0.07 for
SDP. For PD3, SP is 0.155 and SDP is 0.93. In stage 4, as it was expected the FD present
the highest values, having for SP 0.38 and for SDP 0.29. In all stages the FD shows an
performance increase of the SDP when compared to SP for the CoG estimation in different
PD stages. These preliminary results support the role of the spring-damper in the physical
models to simulate pathological gait patterns.
Table 3-1: Frechet distance for spring pendulum and spring damper pendulum pendulum
in the CoG estimation
PD 2 PD 3 PD 4
SP 0.137 0.155 0.38
SDP 0.07 0.09 0.29
Finally table 3-2 shows the parameter values obtained to estimate the CoG in DSP model.
The k and b parameters reflect in a general way the changes produced during the progression
of this pathology. In the model the k value is higher as the disease progresses, and at the same
time b value shows a greatest value in the highest stage that can be interpreted physiologically
as the resistance increment when a movement is produced.
Table 3-2: k and b parameters estimated by stages (SDP)
PD 2 PD 3 PD 4
k 130 135 161.6
b 0.73 1.75 1.83
3.4 Conclusions
This paper presents an performance analysis of the SDP gait model when simulating the
center of gravity trajectory (CoG) for different stages of Parkinson’s disease. Based on the
results obtained with the Frechet distance, in general shows discriminating results, giving to
3.4 Conclusions 21
the SDP a better performance in the 3 stages when it is compared with a baseline method.
Graphical analysis shows the capability of SDP to approximate real trajectories when it
presents amplitude asymmetries. Stage 4 presents the lower similarity results by large CoG
disturbances. Finally, the model parameters (k and b) have a physiological representation
and the value obtained might effectively aid the clinician to quantitatively describe this
pathology .
4 Conclusions and Perspectives
Conclusions
This thesis has explored the problem to quantify gait motor patterns in PD progression in-
cluding new source of information from the upper body. A main contribution has introduced
the trunk bending information to changes in the CoG trajectory, that can be used as a basis
in physical gait models for the simulation and quantification of gait patterns. The integration
of trunk kinematics with gait patterns as the CoG is essential to understand gait alterations,
especially in PD where balance disorders have major impact on functional independence. Ad-
ditionally, the refinement of physical gait models allows an accurate quantitative description
of the disease with physiological interpretability that can support clinicians in the assess-
ment of PD progression and therapeutic interventions. In this research was observed an
inherent problem of Parkinson’s Disease, the high variability between the patients produced
by the underlying motor alterations and compensation mechanisms, what makes modeling a
complex process, but necessary for the disease understanding. Finally, this work contributed
to the building of a dataset of parkinsonian gait that can be useful in the ongoing research
in this field of knowledge.
Perspectives
For health professionals, the possibility of obtaining quantitative and complementary infor-
mation about the patients, refines the medical decision making, resulting in more accurate
treatments that impact directly on the patient quality of-life. Further studies should be
undertaken to achieve the objective of constructing computational methods for the quantifi-
cation of motor patterns in clinical areas and to become an accessible tool in the medical
consultation and rehabilitation fields. The development of this tools will assess the objec-
tively comparison of the information inter e intra patient enabling the better understanding
of the natural history of the disease and the development of strategies to treat the disease
symptoms. Further researches can taking into account issues as: Larger number of subjects,
specially from the first and fourth PD stages, the development of more robust computa-
tional models, that do not require large infrastructure or complex equipments and the the
preparation of these tools for implementation.
Bibliography
[1] Adkin, AL ; Bloem, BR ; Allum, JHJ: Trunk sway measurements during stance and
gait tasks in Parkinson’s disease. En: Gait & posture 22 (2005), Nr. 3, p. 240–249
[2] Allen, Natalie E. ; Schwarzel, Allison K. ; Canning, Colleen G.: Recurrent falls
in Parkinson?s disease: a systematic review. En: Parkinson?s Disease 2013 (2013)
[3] Alves, Guido ; Wentzel-Larsen, Tore ; Aarsland, Dag ; Larsen, Jan P.: Pro-
gression of motor impairment and disability in Parkinson disease a population-based
study. En: Neurology 65 (2005), Nr. 9, p. 1436–1441
[4] Baltadjieva, Rossitza ; Giladi, Nir ; Gruendlinger, Leor ; Peretz, Chava ;
Hausdorff, Jeffrey M.: Marked alterations in the gait timing and rhythmicity of
patients with de novo Parkinson’s disease. En: European Journal of Neuroscience 24
(2006), Nr. 6, p. 1815–1820
[5] Biewener, Andrew A. ; Wakeling, James M. ; Lee, Sabrina S. ; Arnold, Allison S.:
Validation of Hill-Type Muscle Models in Relation to Neuromuscular Recruitment and
Force–Velocity Properties: Predicting Patterns of In Vivo Muscle Force. En: Integrative
and comparative biology (2014), p. icu070
[6] Boonstra, Tjitske A. ; van Vugt, Jeroen P. ; van der Kooij, Herman ; Bloem,
Bastiaan R.: Balance asymmetry in Parkinson?s disease and its contribution to freezing
of gait. (2014)
[7] Camara, Carmen ; Isasi, Pedro ; Warwick, Kevin ; Ruiz, Virginie ; Aziz, Tipu
; Stein, John ; Bakstein, Eduard: Resting tremor classification and detection in
Parkinson’s disease patients. En: Biomedical Signal Processing and Control 16 (2015),
p. 88–97
[8] Capecci, Marianna ; Serpicelli, Chiara ; Fiorentini, Luca ; Censi, Giovanna ;
Ferretti, Matteo ; Orni, Chiara ; Renzi, Rosita ; Provinciali, Leandro ; Cer-
avolo, Maria G.: Postural rehabilitation and kinesio taping for axial postural disorders
in Parkinson’s disease. En: Archives of physical medicine and rehabilitation 95 (2014),
Nr. 6, p. 1067–1075
24 Bibliography
[9] Ceccato, Jean-Charles ; De Seze, Mathieu ; Azevedo, Christine ; Cazalets, Jean-
Rene: Comparison of trunk activity during gait initiation and walking in humans. En:
PLoS One 4 (2009), Nr. 12, p. e8193
[10] Chastan, Nathalie ; Do, Manh C. ; Bonneville, Fabrice ; Torny, Frederic ; Bloch,
Frederic ; Westby, GW ; Dormont, Didier ; Agid, Yves ; Welter, Marie-Laure:
Gait and balance disorders in Parkinson’s disease: impaired active braking of the fall
of centre of gravity. En: Movement Disorders 24 (2009), Nr. 2, p. 188–195
[11] Cifuentes, Christian ; Martinez, Fabio ; Romero, Eduardo: Theoretical and com-
putational analysis of normal and pathological gait: a rewiew. En: Revista Med 18
(2010), Nr. 2, p. 182–196
[12] contrib, Charmant. J. : Kinovea 0.8.15 [computer software]., 2006
[13] Cristani, Marco ; Raghavendra, R ; Del Bue, Alessio ; Murino, Vittorio: Human
behavior analysis in video surveillance: A social signal processing perspective. En:
Neurocomputing 100 (2013), p. 86–97
[14] Damsgaard, Michael ; Rasmussen, John ; Christensen, Søren T. ; Surma, Egidijus
; de Zee, Mark: Analysis of musculoskeletal systems in the AnyBody Modeling System.
En: Simulation Modelling Practice and Theory 14 (2006), Nr. 8, p. 1100–1111
[15] Derrick, Timothy R. ; Caldwell, Graham E. ; Hamill, Joseph: Modeling the
stiffness characteristics of the human body while running with various stride lengths.
En: Journal of Applied Biomechanics 16 (2000), Nr. 1, p. 36–51
[16] Detrembleur, Christine ; Van den Hecke, A ; Dierick, Frederic: Motion of the
body centre of gravity as a summary indicator of the mechanics of human pathological
gait. En: Gait & posture 12 (2000), Nr. 3, p. 243–250
[17] Domınguez, Ana: Aspectos economicos de la enfermedad de Parkinson: coste-
efectividad de tratamientos. En: Rev Esp Econ Salud (2004)
[18] Egerton, Thorlene ; Danoudis, Mary ; Huxham, Frances ; Iansek, Robert: Central
gait control mechanisms and the stride length–cadence relationship. En: Gait & posture
34 (2011), Nr. 2, p. 178–182
[19] Evans, Jonathan R. ; Mason, Sarah L. ; Williams-Gray, Caroline H. ; Foltynie,
Thomas ; Brayne, Carol ; Robbins, Trevor W. ; Barker, Roger A.: The natural
history of treated Parkinson’s disease in an incident, community based cohort. En:
Journal of Neurology, Neurosurgery & Psychiatry 82 (2011), Nr. 10, p. 1112–1118
Bibliography 25
[20] Ferrarin, Maurizio ; Lopiano, L ; Rizzone, M ; Lanotte, M ; Bergamasco, B ;
Recalcati, M ; Pedotti, A: Quantitative analysis of gait in Parkinson’s disease: a
pilot study on the effects of bilateral sub-thalamic stimulation. En: Gait & posture 16
(2002), Nr. 2, p. 135–148
[21] Garcia, Mariano ; Chatterjee, Anindya ; Ruina, Andy ; Coleman, Michael: The
simplest walking model: stability, complexity, and scaling. En: Journal of biomechanical
engineering 120 (1998), Nr. 2, p. 281–288
[22] Gard, Steven A. ; Miff, Steve C. ; Kuo, Arthur D.: Comparison of kinematic and
kinetic methods for computing the vertical motion of the body center of mass during
walking. En: Human movement science 22 (2004), Nr. 6, p. 597–610
[23] Gerus, Pauline ; Rao, Guillaume ; Berton, Eric: Subject-specific tendon-aponeurosis
definition in Hill-type model predicts higher muscle forces in dynamic tasks. (2012)
[24] Geyer, Hartmut ; Seyfarth, Andre ; Blickhan, Reinhard: Spring-mass running:
simple approximate solution and application to gait stability. En: Journal of theoretical
biology 232 (2005), Nr. 3, p. 315–328
[25] Geyer, Hartmut ; Seyfarth, Andre ; Blickhan, Reinhard: Compliant leg behaviour
explains basic dynamics of walking and running. En: Proceedings of the Royal Society
of London B: Biological Sciences 273 (2006), Nr. 1603, p. 2861–2867
[26] Goldberg, Allon ; Hernandez, Manuel E. ; Alexander, Neil B.: Trunk repo-
sitioning errors are increased in balance-impaired older adults. En: The Journals of
Gerontology Series A: Biological Sciences and Medical Sciences 60 (2005), Nr. 10, p.
1310–1314
[27] Grabli, David ; Karachi, Carine ; Welter, Marie-Laure ; Lau, Brian ; Hirsch,
Etienne C. ; Vidailhet, Marie ; Francois, Chantal: Normal and pathological gait:
what we learn from Parkinson’s disease. En: Journal of Neurology, Neurosurgery &
Psychiatry 83 (2012), Nr. 10, p. 979–985
[28] Hamner, Samuel R. ; Seth, Ajay ; Steele, Katherine M. ; Delp, Scott L.: A rolling
constraint reproduces ground reaction forces and moments in dynamic simulations of
walking, running, and crouch gait. En: Journal of biomechanics 46 (2013), Nr. 10, p.
1772–1776
[29] Hiroyuki Hori, MD: A systematic review of the worldwide prevalence and incidence
of Parkinson?s disease. En: J Med Assoc Thai 94 (2011), Nr. 6, p. 749–55
[30] Hong, Hyunhwa ; Kim, Seyoung ; Kim, Cheolwoong ; Lee, Soonhyuck ; Park,
Sukyung: Spring-like gait mechanics observed during walking in both young and older
adults. En: Journal of biomechanics 46 (2013), Nr. 1, p. 77–82
26 Bibliography
[31] Horsman, MD K. ; Koopman, HFJM ; Van der Helm, FCT ; Prose, L P. ;
Veeger, HEJ: Morphological muscle and joint parameters for musculoskeletal mod-
elling of the lower extremity. En: Clinical biomechanics 22 (2007), Nr. 2, p. 239–247
[32] Ijspeert, Auke J.: Central pattern generators for locomotion control in animals and
robots: a review. En: Neural Networks 21 (2008), Nr. 4, p. 642–653
[33] Jankovic, Joseph: Parkinson?s disease: clinical features and diagnosis. En: Journal
of Neurology, Neurosurgery & Psychiatry 79 (2008), Nr. 4, p. 368–376
[34] Jiuzhou, Zhao ; Hehua, Ju: Detail analysis on human bipedalism and a natural gait
generation method for biped robots. En: Mechatronics and Automation (ICMA), 2013
IEEE International Conference on IEEE, 2013, p. 873–878
[35] Jurcevic Lulic, T ; Muftic, O [u. a.]: Trajectory of the human body mass centre
during walking at different speed. En: DS 30: Proceedings of DESIGN 2002, the 7th
International Design Conference, Dubrovnik, 2002
[36] Kim, Samuel D. ; Allen, Natalie E. ; Canning, Colleen G. ; Fung, Victor S.: Postural
instability in patients with Parkinson?s disease. En: CNS drugs 27 (2013), Nr. 2, p.
97–112
[37] Kim, Seyoung ; Park, Sukyung: Leg stiffness increases with speed to modulate gait
frequency and propulsion energy. En: Journal of Biomechanics 44 (2011), Nr. 7, p.
1253–1258
[38] Kowal, Stacey L. ; Dall, Timothy M. ; Chakrabarti, Ritashree ; Storm,
Michael V. ; Jain, Anjali: The current and projected economic burden of Parkinson’s
disease in the United States. En: Movement Disorders 28 (2013), Nr. 3, p. 311–318
[39] Kucinski, Aaron ; Albin, Roger L. ; Lustig, Cindy ; Sarter, Martin: Modeling
falls in Parkinson’s disease: Slow gait, freezing episodes and falls in rats with extensive
striatal dopamine loss. En: Behavioural brain research 282 (2015), p. 155–164
[40] Lee, Cynthia R. ; Farley, Claire T.: Determinants of the center of mass trajectory
in human walking and running. En: The Journal of Experimental Biology 201 (1998),
Nr. 21, p. 2935–2944
[41] Linley, Heather S. ; Sled, Elizabeth A. ; Culham, Elsie G. ; Deluzio, Kevin J.:
A biomechanical analysis of trunk and pelvis motion during gait in subjects with knee
osteoarthritis compared to control subjects. En: Clinical Biomechanics 25 (2010), Nr.
10, p. 1003–1010
[42] Lotharius, Julie ; Brundin, Patrik: Pathogenesis of Parkinson’s disease: dopamine,
vesicles and α-synuclein. En: Nature Reviews Neuroscience 3 (2002), Nr. 12, p. 932–942
Bibliography 27
[43] Maetzler, Walter ; Liepelt, Inga ; Berg, Daniela: Progression of Parkinson’s
disease in the clinical phase: potential markers. En: The Lancet Neurology 8 (2009),
Nr. 12, p. 1158–1171
[44] Maetzler, Walter ; Mancini, Martina ; Liepelt-Scarfone, Inga ; Muller, Katha-
rina ; Becker, Clemens ; Van Lummel, Rob C. ; Ainsworth, Erik ; Hobert,
Markus ; Streffer, Johannes ; Berg, Daniela [u. a.]: Impaired trunk stability in
individuals at high risk for Parkinson?s disease. En: PLoS One 7 (2012), Nr. 3, p.
e32240
[45] Magnenat-Thalmann, Nadia ; Ratib, Osman ; Choi, Hon F.: 3D Multiscale Phys-
iological Human. Springer, 2014
[46] Mancini, Martina ; Horak, Fay B. ; Zampieri, Cris ; Carlson-Kuhta, Patricia ;
Nutt, John G. ; Chiari, Lorenzo: Trunk accelerometry reveals postural instability in
untreated Parkinson?s disease. En: Parkinsonism & related disorders 17 (2011), Nr. 7,
p. 557–562
[47] Marti-Puig, Pere ; Rodrıguez, Sara ; De Paz, Juan F. ; Reig-Bolano, Ramon ;
Rubio, Manuel P. ; Bajo, Javier: Stereo video surveillance multi-agent system: new
solutions for human motion analysis. En: Journal of Mathematical Imaging and Vision
42 (2012), Nr. 2-3, p. 176–195
[48] Martınez, Fabio ; Cifuentes, Christian ; Romero, Eduardo: Simulation of normal
and pathological gaits using a fusion knowledge strategy. En: Journal of neuroengineer-
ing and rehabilitation 10 (2013), Nr. 1, p. 73
[49] Martinez, Fabio ; Gomez, Francisco ; Romero, Eduardo: A kinematic method for
computing the motion of the body centre-of-mass (CoM) during walking: a Bayesian
approach. En: Computer methods in biomechanics and biomedical engineering 14
(2011), Nr. 06, p. 561–572
[50] Mateus, Ceu ; Coloma, Joana: Health Economics and Cost of Illness in Parkinson?s
Disease. En: European Neurological Review 8 (2013), p. 6–9
[51] Matinolli, M ; Korpelainen, JT ; Sotaniemi, KA ; Myllyla, VV ; Kor-
pelainen, R: Recurrent falls and mortality in Parkinson?s disease: a prospective
two-year follow-up study. En: Acta Neurologica Scandinavica 123 (2011), Nr. 3, p.
193–200
[52] Moore, Steven T. ; MacDougall, Hamish G. ; Gracies, Jean-Michel ; Cohen,
Helen S. ; Ondo, William G.: Long-term monitoring of gait in Parkinson’s disease. En:
Gait & Posture 26 (2007), Nr. 2, p. 200–207
28 Bibliography
[53] Muller, J ; Wenning, GK ; Jellinger, K ; McKee, A ; Poewe, W ; Litvan,
I: Progression of Hoehn and Yahr stages in Parkinsonian disorders: a clinicopathologic
study. En: Neurology 55 (2000), Nr. 6, p. 888–891
[54] Nadeau, Sylvie ; Amblard, Bernard ; Mesure, Serge ; Bourbonnais, Daniel:
Head and trunk stabilization strategies during forward and backward walking in healthy
adults. En: Gait & posture 18 (2003), Nr. 3, p. 134–142
[55] Nikooyan, Ali A. ; Zadpoor, Amir A.: Mass–spring–damper modelling of the hu-
man body to study running and hopping-an overview. En: Proceedings of the Institu-
tion of Mechanical Engineers, Part H: Journal of Engineering in Medicine (2011), p.
0954411911424210
[56] Nutt, JG ; Marsden, CD ; Thompson, PD: Human walking and higher-level gait
disorders, particularly in the elderly. En: Neurology 43 (1993), Nr. 2, p. 268–268
[57] Peppe, A ; Chiavalon, C ; Pasqualetti, P ; Crovato, D ; Caltagirone, C: Does
gait analysis quantify motor rehabilitation efficacy in Parkinson’s disease patients? En:
Gait & posture 26 (2007), Nr. 3, p. 452–462
[58] Pfeiffer, R.F. ; Wszolek, Z.K. ; Ebadi, M.: Parkinson’s Disease, Second Edition.
Taylor & Francis, 2012. – ISBN 9781439807149
[59] de Pina Filho, Armando C. ; Dutra, Max S.: Dynamical analysisof a coupled
oscillators system used for human gait simulation. En: Proceedings of COBEM (2011)
[60] Plotnik, Meir ; Giladi, Nir ; Dagan, Yaacov ; Hausdorff, Jeffery M.: Postural
instability and fall risk in Parkinson?s disease: impaired dual tasking, pacing, and
bilateral coordination of gait during the ?ON? medication state. En: Experimental
brain research 210 (2011), Nr. 3-4, p. 529–538
[61] Rodriguez-Oroz, Maria C. ; Jahanshahi, Marjan ; Krack, Paul ; Litvan, Irene
; Macias, Raul ; Bezard, Erwan ; Obeso, Jose A: Initial clinical manifestations
of Parkinson’s disease: features and pathophysiological mechanisms. En: The Lancet
Neurology 8 (2009), Nr. 12, p. 1128–1139
[62] Schrag, Anette ; Choudhury, Mahbuba ; Kaski, Diego ; Gallagher, David A.:
Why do patients with Parkinson?s disease fall? A cross-sectional analysis of possible
causes of falls. En: npj Parkinson’s Disease 1 (2015)
[63] Seth, Ajay ; Sherman, Michael ; Reinbolt, Jeffrey A. ; Delp, Scott L.: OpenSim:
a musculoskeletal modeling and simulation framework for in silico investigations and
exchange. En: Procedia Iutam 2 (2011), p. 212–232
Bibliography 29
[64] Sofuwa, Olumide ; Nieuwboer, Alice ; Desloovere, Kaat ; Willems, Anne-Marie
; Chavret, Fabienne ; Jonkers, Ilse: Quantitative gait analysis in Parkinson?s
disease: comparison with a healthy control group. En: Archives of physical medicine
and rehabilitation 86 (2005), Nr. 5, p. 1007–1013
[65] Suri, Pradeep ; Kiely, Dan K. ; Leveille, Suzanne G. ; Frontera, Walter R. ;
Bean, Jonathan F.: Trunk muscle attributes are associated with balance and mobility
in older adults: a pilot study. En: PM&R 1 (2009), Nr. 10, p. 916–924
[66] Tassorelli, C ; De Icco, R ; Alfonsi, E ; Bartolo, M ; Serrao, M ; Avenali, M
; De Paoli, I ; Conte, C ; Pozzi, NG ; Bramanti, P [u. a.]: Botulinum toxin type
A potentiates the effect of neuromotor rehabilitation of Pisa syndrome in Parkinson
disease: A placebo controlled study. En: Parkinsonism & related disorders 20 (2014),
Nr. 11, p. 1140–1144
[67] Voon, Valerie ; Gao, Jennifer ; Brezing, Christina ; Symmonds, Mkael ;
Ekanayake, Vindhya ; Fernandez, Hubert ; Dolan, Raymond J. ; Hallett, Mark:
Dopamine agonists and risk: impulse control disorders in Parkinson’s; disease. En:
Brain 134 (2011), Nr. 5, p. 1438–1446
[68] Waters, Robert L. ; Mulroy, Sara: The energy expenditure of normal and pathologic
gait. En: Gait & posture 9 (1999), Nr. 3, p. 207–231
[69] Wren, Tishya A. ; Gorton, George E. ; Ounpuu, Sylvia ; Tucker, Carole A.:
Efficacy of clinical gait analysis: A systematic review. En: Gait & posture 34 (2011),
Nr. 2, p. 149–153
[70] Zhao, Ying J. ; Wee, Hwee L. ; Chan, Yiong-Huak ; Seah, Soo H. ; Au, Wing L. ;
Lau, Puay N. ; Pica, Emmanuel C. ; Li, Shu C. ; Luo, Nan ; Tan, Louis: Progression
of Parkinson’s disease as evaluated by Hoehn and Yahr stage transition times. En:
Movement Disorders 25 (2010), Nr. 6, p. 710–716