See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/221976076

Gait measures with a triaxial accelerometer among patients with neurological

impairment

Article in Neurological Sciences · March 2012

DOI: 10.1007/s10072-012-1017-x · Source: PubMed

CITATIONS READS

28 105

8 authors, including:

Patrik Fazio Gino Granieri

Karolinska Institutet University of Ferrara
36 PUBLICATIONS 564 CITATIONS 4 PUBLICATIONS 137 CITATIONS

SEE PROFILE SEE PROFILE

Ilaria Casetta Sante Mazzacane

University of Ferrara University of Ferrara
149 PUBLICATIONS 3,471 CITATIONS 43 PUBLICATIONS 222 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Solvent exposure and MS View project

Sleep and stroke View project

All content following this page was uploaded by Sante Mazzacane on 03 May 2019.

The user has requested enhancement of the downloaded file.

Neurol Sci
DOI 10.1007/s10072-012-1017-x
ORIGINAL ARTICLE
Gait measures with a triaxial accelerometer among patients
with neurological impairment
Patrik Fazio • Gino Granieri • Ilaria Casetta •
Edward Cesnik • Sante Mazzacane • Pietro Caliandro
Francesco Pedrielli • Enrico Granieri
Received: 6 March 2012 / Accepted: 7 March 2012
Ó Springer-Verlag 2012
Abstract The purpose of the present study is to evaluate
accelerometric parameters of gait in different neurological
conditions with pathological gait impairment compared to
healthy subjects. We studied 17 patients affected by Par-
kinson’s disease, 24 with ataxic gait due to different dis-
eases and 24 healthy subjects supplied with a triaxial
accelerometer with a portable datalogger which measures
acceleration and deceleration on an anterior–posterior,
mediolateral and vertical plane at an approximate level of
the center of mass (back sacral localization) and in other
two positions (sternal and frontal sacral region) during a
steady-state walking. Analyses of the basic accelerometric
parameters associated with a jerk analysis allowed us to
differentiate between the population groups. We observed a
significant reduction of acceleration parameters in neuro-
logical patients when compared with healthy subjects, with
a reduction of the mean acceleration of 0.30 m/s2 for ataxic
P. Fazio
I. Casetta
E. Cesnik
P. Caliandro
E. Granieri
Department of Medical, Surgical Sciences of Communication
and Behavior, Section of Neurology, University of Ferrara,
Corso della Giovecca 203, 44100 Ferrara, Italy
P. Fazio (&)
Department of Neurophysiology and Epilepsy, Foundation
IRCCS Neurological Institute ‘‘C. Besta’’ (INNCB), Milan, Italy
e-mail: patrik.fazio@unife.it
G. Granieri
Group ProMot, Department of Medical, Surgical Sciences
of Communication and Behavior, Section of Neurology,
University of Ferrara, Ferrara, Italy
S. Mazzacane
Department of Architecture, University of Ferrara, Ferrara, Italy
F. Pedrielli
Department of Physics, University of Ferrara, Ferrara, Italy
•
and 0.64 m/s2 for parkinsonian patients (t test, p \ 0.01).
The root-mean square of the accelerations was used to
quantify the attenuations of accelerations. This study sug-
gests that a triaxial accelerometer is a good practical and an
economic tool for assessing the alteration of perambula-
tion. Moreover, it is plausible to use these data to obtain
objective parameters in the evaluation of the progression of
the disease and the efficacy of therapeutic tools.
Keywords Gait analysis
Accelerometry

Triaxial accelerometer
Parkinson’s disease
Ataxia
Introduction
Many disorders of motor function are most clearly mani-
fested as impairment of gait. The clinical evaluation of
patients with gait impairment is mostly subjective and the
detection of abnormalities in terms of posture, cadence,
step length and base commonly forms opinions or
impressions only useful for diagnostic purposes [1–6]. In
terms of analysis of these data, for their reproducibility and
their use in a monitoring context (progression of disease or
response to treatment), it is clearly preferable to measure
objectively the walking pattern [6–8].
Attention is growing to gather variables with instru-
ments that can quantify the aspects of gait such as camera-
assisted motion analysis, pedometers, electromyography
for kinesiology studies and opto-electronic systems for
kinematics and accelerometers [8–12]. Unfortunately, the
majority of these techniques need expensive measuring
instruments. In the present study, we decided to use
accelerometers to obtain, with a simple acquisition system,
a complementary and objective parameter to add to the
clinical scales and other evaluations. Accelerometers are
123

sensors that provide electric voltages proportional the
accelerations of a mechanical transducer. The ones adopted
in this work are based on three very small spring–mass,
inertial–elastic transducers, oriented orthogonally to each
other. The acceleration signal is elaborated to provide
velocity changes (through integration) and/or acceleration
changes (so called ‘‘jerk’’; by derivation), thus giving an
elaborated picture of the body motion, given that only
changes of speed are detected and information on the initial
status is missing [13, 14]. In relation to the present study,
the accelerations experienced by the trunk, during steady-
state walking, reflect the accelerations imposed on the body
center of gravity by the pendulum-like mechanism of gait
in both the sagittal and the frontal plane. This entails
positive and negative accelerations, both in the forward and
the upward direction [15]. The repeated patterns obtained
during walking with different linear accelerometers located
in different positions of the body could contain information
on the smoothness or variability of the walking pattern
(index of smoothness), especially by using jerk (time
derivative of acceleration) analysis, as previously reported
by other authors [13, 14, 22, 23]. One of the benefits of
using accelerometers, compared to more traditional gait
analysis instruments, is their low cost and testing is not
restricted to a laboratory environment; accelerometers are
small and this allow a ‘‘natural’’ gait without any limitation
due to instrumentation [17]. The method for the measure-
ment of acceleration is not new: a series of papers by
Moe-Nilsson [16, 17] described the use of triaxial accel-
erometers to measure walking stability and the output from
accelerometry is reported as a stable indicator of overall
body movements. In 2003, Paquet et al. [18] collected a
series of observations on gait on PD patients and their
accelerometric measures. In their study, they observed a
reduction in the cranial–caudal activity which was inter-
preted as a sign of hypokinesia in parkinsonian patients.
More recently, Mancini et al. [19] evaluated the postural
sway with a linear accelerometer on posterior trunk, and it
was used as a reliable biomarker to evaluate postural sta-
bility in early Parkinson’s disease during the progression of
the disease. In people with movement disorders, accelero-
metric measures could represent a complementary choice
for the evaluation of the variability of movements and
Table 1 Age distribution of subjects under investigation for gait analysis: healthy subjects, Parkinson’s disease patients and ataxic patients
Age groups
20–39 40–49
Healthy subjects 8 6
Patients with Parkinson’s disease – –
Ataxic patients 7 6
123
Neurol Sci
balance, providing a non-invasive, economic and portable
method for the measurements. The purpose of the present
study was to identify and describe, with a simple method,
the ‘‘accelerometric imprint’’ between different neurolog-
ical conditions implying gait disorders compared with a
healthy control population.
Materials and methods
Subjects
Gait analysis was conducted in the Neurological Clinic of
the University Hospital of Ferrara, Italy, where studies on
Promotion of Motor Activity in Neurological Disorders are
in progress. Sixty-five subjects took part in the experiment
after giving their informed consent. Twenty-four were
healthy subjects without gait impairments (12 men and 12
women), 17 were patients affected by Parkinson’s disease
(11 men and 6 women) and, with a mean Unified Parkin-
son’s Disease Rating Scale (UPDRS III), 22.5 ± 3.6, 24
were ataxic patients (13 men and 11 women). The ataxic
patients included in the study were central ataxic patients,
namely caused by multiple sclerosis (mean EDSS 4): one
case of parietal stroke, one case of chronic encephalopathy
and one case of gait apraxia. For the age distribution of the
enrolled subjects, see Table 1. Parkinsonian patients were
tested approximately 1 h after their midmorning dose of
medication, during the ‘‘on phase’’, to avoid fluctuations of
the gait pattern induced by levodopa [20].
Materials and procedure
The technique we used is based on the recording of the
accelerations in the median sacral region, in front of and
behind S2, both landmarks approaching horizontal pro-
jections of the center of body mass [21, 22], and in front of
the sternum (Fig. 1). We used a triaxial piezoelectric
accelerometer tied to the body by an elastic belt. The
device had a sampling rate of 20 Hz. The average output
from subsequent 0.05 s time frames was transmitted to the
personal computer via Bluetooth technology, which chops
up the data being sent and transmits in the range
Total
50–59 60–69 70–85
4 3 3 24
– 5 12 17
2 5 4 24
Neurol Sci
Fig. 1 Triaxial piezoelectric accelerometer and standard positions in
sternal region 1, frontal sacral region 2 and back sacral region 3 for
recording of accelerations
2,400–2,483.5 MHz. Data were stored on a data recorder
for successive analyses. The accelerometer has a 0.006 m/s2
accuracy. Accelerometers measure the instant accelerations
as a vectorial sum of the three axial components after the
subtraction of the constant quantity. During the whole
experiment, patients were observed by the evaluator, and
the subject’s performance was videotaped.
Walking test
We measured median acceleration and speed in three
different conditions: An accelerometer was tied to the
sternal region (sternal region, SR), the second one was tied
to the front part of the sacral region (frontal sacral, FS)
and the third one behind the sacral region (back sacral,
BS) where the body center of mass (CoM) is located in a
quiet standing position [21, 22]. The subject, equipped
with the accelerometer, was asked to walk on average
three times (one for every experimental condition) forth
and back at a comfortable speed for him/her, in a 25-m
long straight corridor. We chose to make subjects walk at
their comfortable, relaxed speed because it was supposed
to be the optimal condition for comparative studies.
Moreover, while walking, each individual has his own
preferred combination of speed, pace and step length, and
this self-selected walking pattern is also preferable in
relation to maintain stability during gait (steady-state
walking speed).
A straight 25-m-long corridor can reproduce a period of
stabilized walking that is longer than 20 steps, a condition
that is recognized to improve accuracy in studies of gait
disorders [18, 19]. For the statistical analysis, the distance
of 25 m allowed us to discard data from non-stabilized
walk (beginning and end of a walking test).
Data analysis
For every experimental condition, we chose data from the
steady part of the walk, eliminating the beginning and the
end of the test where the subjects accelerated and decel-
erated, respectively, on the anterior–posterior axis. We
computed the average walking speeds by dividing the
distance by the time each patient it took (except from the
column of the time in the acceleration trace) and as a
supplementary control we multiplied 0.05 s for the total
amount of points in raw traces where there were move-
ments. For the accelerations (vectorial sum of the three
axial components), we computed the quantity for every
position of the accelerometer (SR, FS and BS) of each
healthy (H), Parkinson’s disease (PD) and ataxic (A) sub-
ject. The analysis of variance (ANOVA), with a post hoc
Scheffé’s test and the t test, was used to compare means.
Jerk analysis (i.e., the derivatives of the acceleration) was
calculated from the resultant of acceleration (Acc) mea-
sured at thoracic level (SR). The root mean square of the
accelerations (RMS) was used to define the magnitude and
trends of acceleration traces with absolute and mean values
for each category of subjects.
Results
Mean walking speed for healthy subjects was 1.64 m/s (SD
0.27), for PD patient’s 0.88 m/s (SD 0.25) and for the
ataxic group 0.94 m/s (SD 0.32). The analysis for mean
acceleration over all the positions during steady-state walk
showed that normal subjects had higher mean values than
ataxic patients (p \ 0.001) and PD patients (p \ 0.001).
Healthy subjects showed a mean acceleration of 1.38 m/s2,
while ataxic patients had an average of 1.07 m/s2 and PD
patients an average of 0.74 m/s2 (Table 2).
PD and ataxic patients showed a mean acceleration that
was 23 and 46% lower than that of healthy subjects, and
the difference between the three subgroups of subjects (H,
PD or A) was statistically significant (p \ 0.001). Ataxic
patients’ mean acceleration was globally higher than PD
subjects (p \ 0.001). The data are summarized in Fig. 2
and Table 2. In the healthy control population, sternal
acceleration was lower than the frontal sacral (p \ 0.001)
and posterior sacral acceleration (p \ 0.001). There were
no statistically significant differences between anterior and
posterior sacral acceleration (p = 0.3), although the
acceleration recorded from the frontal sacral site was
greater than the acceleration from the posterior sacral
region. In PD patients, sternal acceleration was signifi-
cantly lower than the one recorded from the anterior sacral
region (p \ 0.0001) and from the posterior sacral region
(p \ 0.0001). Similar results have been reported from
123

Table 2 Mean acceleration (m/s2) ? standard error on the three axes according to type of subject and position of the accelerometer in the body
Fig. 2 Mean acceleration during walking test according to type
(healthy, Parkinson’s disease or ataxic) and site of recording
ataxic patients with an average acceleration from the
sternal region, which is significantly lower than the one
recorded from the anterior sacral region (p \ 0.001) and
from the posterior sacral region (p \ 0.05), without sta-
tistically significant differences between the anterior and
posterior sacral region (p = 0.8). A significant difference
was also found between the frontal and the posterior sacral
region (p \ 0.001). The ratio between the accelerations
acquired from the sternal region and the accelerations taken
from the back sacral region was different within the type of
subjects: 0.75 for healthy subjects, 0.82 for patients with
Parkinson’s disease and 0.69 for ataxic patients. Regarding
the jerk analysis, as expected for acceleration and speed
differences, our healthy groups showed higher jerk values
compared to the PD and ataxic groups. The jerk RMS
values measured at the sternal thoracic level were signifi-
cantly lower in the ataxic group with a jerk RMS 2.108 of
m/s3 with an SD 0.76 and PD patients with a jerk RMS of
2.176 m/s3 with an SD 0.6 than in the healthy group that
showed a Jerk RMS of 3.080 m/s3 with an SD 0.457 (see
Fig. 3).
Discussion and conclusions
Walking is a voluntary process, but stride-to-stride regu-
lation of gait is controlled by the nervous system in a lar-
gely automatic way. Disruption of any portion of the neural
123
Neurol Sci
4 JERK RMS(m/sec3)
Healthy
3.5
Parkins
3
Ataxic
2.5
1.5
0.5
0
Healthy Parkins Ataxic
Fig. 3 Jerk root mean-square values during steady-state walking
according to the different gait disorders (healthy, Parkinson’s disease
or ataxic) at thoracic level
networks involved in motor control can produce uncoor-
dinated and poorly controlled movements. In many cases,
the patient’s gait is a signal of his clinical situation [1]. The
results of this study indicate that accelerometry appears to
have a considerable potential as a measure of walking
stability and smoothness, although its validity is yet to be
fully evaluated in a large number of subjects. Studies on
parkinsonian and ataxic patients’ gait measures are scarce.
Obtaining a global marker of a subject’s gait dynamics was
our aim when we verified that there were differences
detected with acceleration-based parameters acquired from
strategic regions of the body among healthy subjects,
subjects with Parkinson’s disease and ataxic subjects. We
noticed that the mean acceleration values acquired from
every part of the body (SR, FS, BS) for every subject with
PD were different (with a statistical significance) with
respect to the ones acquired from healthy subjects. In this
way, healthy subjects can be distinguished from patients
with PD by using the accelerations acquired from every
part of the body. These findings are very important, espe-
cially because in our sample we had patients with PD who
were clinically well compensated by the therapy and had an
almost macroscopically normal gait. The method we sug-
gest seems to have many advantages for a clinical gait
analysis as reported by Kavanagh [24]. In healthy subjects,
according to the physiology of gait, we found a gradient of
acceleration with maximum values at foot level and min-
imum values at head level that was consistent with those
of previous investigations [25, 26]. Ground reactions,
Neurol Sci
originating under the foot, causes torques around segmental
hinges (the joints) and accelerations are thus smoothed
along a caudal–cranial gradient, due to inertial, elastic and
viscous dampening offered by the various body segments
involved [27, 28]. This attenuation was clearly evident by
the resultant decrease of acceleration going from the pelvis
to the head level. This phenomenon appeared to be
enhanced in our parkinsonian and ataxic patients during
steady-state walk. The reduced global accelerometer values
of patients (differentiating PD patients and ataxic ones)
could reflect only partially the differences showed by
walking speeds in our pathological group, but we could
hypothesize also a maladaptive and compensatory axial
structural body behavior due to the gait alteration induced
by their pathology. This effect was more evident in PD
patients and was confirmed by the jerk analysis. It has been
shown that the time-integrated squared jerk decreases with
increased smoothness of movement for PD and ataxic
groups. Ataxic and Parkinson jerk analysis results con-
stantly showed that acceleration and its variations were
partly independent of the sequence of steps or changing
pitch. In our pathological groups (PD and ataxic), it was
evident that the jerk RMS values were reduced and more
stationary (without specific peaks or high excursion)
probably in relation with the global rigidity. Moreover,
bradykinesia was among the main characteristics of gait in
parkinsonian patients, the best variable for separating
patients from controls. As previously reported, bradykine-
sia stems from progressive reduction of mobility along a
caudo-cranial gradient and may be suitable for quantifying
physical activity in patients [18, 23, 24]. According to
literature and clinical observations, our patients with Par-
kinson disease showed a severely reduced walking speed
[1]. Considering the results of this investigation, the
accelerometer evaluations could be used and proposed as a
simple complementary index of gait stability during the
follow-up of patients with movement disorders, especially
when the results of pharmacological, surgical or rehabili-
tative treatments need to be quantified. The validity of
these findings is yet to be fully evaluated in a large number
of subjects and possibly compared with other techniques.
Nowadays, more complex methods allow estimating the
motion of the center of body mass, by inferring it from the
motion of body segments, through computer-intensive
modeling [29]. In principle, these measurements rely on
inferences heavier than the ones underlying accelerometry
and their practical superiority is not ascertained. Recently,
a conceptually different alternative has been offered by the
direct measurement (not an estimate) of the motion of the
body center of mass, deducted (so called ‘‘Newtonian’’
method) from the ground reactions originated by force
platforms or force treadmill on which the subject can
perform entire successive steps. A recent paper evidenced
that sharp changes of tangential acceleration of the center
of mass may mark step phases (e.g., left–right direction
changes) where balance may be most challenged [30].
These methods are not accessible to routine clinical
assessment, yet their results confirm the promising role of
accelerometry and the results of the present study, in
clinical evaluation of gait.
Acknowledgments The study was supported by Fondazione Cassa
di Risparmio di Cento. The authors thank Dr. Eva Sjolin for her help
in revising the translation of the manuscript. We would also like to
express our sincere appreciation to Mario Manca for helpful com-
ments on the manuscript.
References
1. Knutsson E (1972) An analysis of Parkinsonian gait. Brain
95(3):475–486
2. Kurtzke JF, Beebe GW, Nagler B et al (1977) Studies on the
natural history of multiple sclerosis. Early prognostic features of
the later course of the illness. J Chronic Dis 30:819–830
3. Ramaker C, Marinus J, Stiggelbout AM, Van Hilten BJ (2002)
Systematic evaluation of rating scales for impairment and dis-
ability in Parkinson’s disease. Mov Disord 17:867–876
4. Marsden CD, Schachter M (1981) Assessment of extrapyramidal
disorders. Br J Clin Pharmacol 11:129–151
5. Nutt JG, Marsden CD, Thompson PD (1993) Human walking and
higher-level gait disorders, particularly in the elderly. Neurology
43:268–279
6. Winters JM, Loeb G (2000) Biomechanics and neural control of
posture and movement. Springer, New York
7. Carletti T, Fanelli D, Guarino A (2006) A new route to nonin-
vasive diagnosis in neurodegenerative diseases? Neurosci Lett
394:252–255
8. Busse ME, Pearson OR, van Deursen R, Wiles CM (2004)
Quantified measurement of activity provides insight into motor
function and recovery in neurological disease. J Neurol Neuro-
surg Psychiatry 75:884–888
9. Pélissier J, Boisson D (1994) Les paramètres de la marche humaine
: techniques actuelles d’exploration. In: Pélissier J, Brun V (eds) La
marche humaine et sa pathologie. Masson, Paris, pp 41–55
10. Van Hilten JJ, Middelkoop HAM, Kerkhof GA, Roos RAC
(1991) The assessment of motor activity in Parkinson’s disease: a
new approach. J Neurol Neurosurg Psychiatry 54:976–979
11. Van Hilten JJ, Hoogland G, Van der Velde EA et al (1993)
Quantitative assessment of parkinsonian patients by continuous
wrist activity monitoring. Clin Neuropharmacol 16:36–45
12. Katayama S (2001) Actigraph analysis of diurnal motor fluctua-
tions during dopamine agonist therapy. Eur Neurol 46(Suppl
1):11–17
13. Culhane KM, O’Connor M, Lyons D, Lyons GM (2005) Accel-
erometers in rehabilitation medicine for older adults. Age Ageing
34:556–560
14. Hogan N, Sternad D (2009) Sensitivity of smoothness measures
to movement duration, amplitude, and arrests. J Mot Behav
41(6):529–534
15. Cavagna GA, Tesio L, Fuchimoto T, Heglund NC (1983) Ergo-
metric evaluation of pathological gait. J Appl Physiol 55(2):607–
613
16. Moe-Nilssen R (1998) A new method for evaluating motor control
in gait under real-life environmental conditions. Part 1 and 2: the
instrument and gait analysis. Clin Biomech 13:320–335
123

17. Moe-Nilssen R (1998) Test–retest reliability of trunk acceler-
ometry during standing and walking. Arch Phys Med Rehabil
79:1377–1385
18. Paquet JM, Auvinet B, Chaleil D, Barrey E (2003) Analyse des
troubles de la marche par une methode accélérométrique dans la
maladie de Parkinson. Rev Neurol (Paris) 159:768–789
19. Mancini M, Horak FB, Zampieri C, Carlson-Kuhta P, Nutt JG,
Chiari L (2011) Trunk accelerometry reveals postural instability
in untreated Parkinson’s disease. Parkinsonism Relat Disord
17(7):557–562
20. Caliandro P, Ferrarin M, Cioni M, Bentivoglio AR, Minciotti I
et al (2011) Levodopa effect on electromyographic activation
patterns of tibialis anterior muscle during walking in Parkinson’s
disease. Gait Posture 33(3):436–441
21. Winter DA (1991) The biomechanics and motor control of human
gait: normal, elderly and pathological, 2nd edn. University of
Waterloo Press, Waterloo
22. Winter DA (1995) Human balance and posture control during
standing and walking. Gait Posture 3:193–214
23. Kamen G, Patten C, Sison S (1998) An accelerometry-based
system for the assessment of balance and postural sway. Geron-
tology 44:40–45
123
View publication stats
Neurol Sci
24. Kavanagh JJ, Hylton BM (2008) Accelerometry: a technique for
quantifying movement patterns during walking. Gait Posture
28:1–15
25. Hirasaki E, Kubo T, Nozawa S, Matano S, Matsunaga T (1993)
Analysis of head and body movements of elderly people during
locomotion. Acta Otolaryngologica 501:25–30
26. Smeathers JE (1989) Measurement of transmissibility for the
human spine during walking and running. Clin Biomech 4:34–40
27. Pozzo T, Berthoz A, Lefort L (1990) Head stabilization during
various locomotor tasks in humans. I. Normal subjects. Exp Brain
Res 82:97–106
28. Kavanagh JJ, Morrison S, Barrett RS (2005) Coordination of
head and trunk accelerations during walking. Eur J Appl Physiol
94(4):468–475
29. Orendurff MS, Segal AD, Klute GK, Berge JS, Rohr ES, Kadel
NJ (2004) The effect of walking speed on center of mass dis-
placement. J Rehabil Res Dev 41(6):829–834
30. Tesio L, Rota V, Perucca L (2011) The 3D trajectory of the body
centre of mass during adult human walking: evidence for a speed-
curvature power law. J Biomech 24(44):732–740