Influence of Audio Biofeedback on Structural
Properties of Postural Sway
L. Chiari1, M. Dozza1,2, A. Cappello1, F.B. Horak2
OHSU1Dipartimento di Elettronica, Informatica e Sistemistica
Alma Mater Studiorum - Università di Bologna, Italia2Neurological Sciences Institute
OHSU, Beaverton (OR), USA
What is Audio Biofeedback (ABF)?
(van der Kooij, 2000)
BALANCE
SENSES
MUSCLESBRAIN
Sensory Integration
InternalMap
• Balance is the consequence of appropriate muscle activations processed by the brain fusion of sensory information
BALANCE
MUSCLESBRAIN
Sensory Integration
InternalMap
• Visual, Vestibular and Somatosensory information are the major cues used by the brain to perform balance
VISION
VESTIBULAR
SOMATOS.
AUDITORY
• ABF may be used to involve more largely in the “game” the AUDITORY channel
Diapason: a recent solution for ABF
(Chiari et al., IEEE Trans Biomed Eng, submitted)
Sensor characteristics
• The sensor used is able to provide the complete linear & angular kinematics of the trunk (3 accelerometers, 3 gyroscopes)
• ABF in its present release uses only 2-D acceleration (AP and ML directions)
(Giansanti et al., Proc. ISPG, Maastricht, 2001)
Accelerometricsensors
Amplifier andlow-pass filter
SENSOR BOX
TRANSMISSIONBOX
Release 1 - 2002
Diapason: the sonification procedure
Safety Region (SR)
• represents the limit of stability
• is the region in which the COM projection is inside the subject’s support base
• the support base is processed on anthropometric parameters (feet length and width)
Reference Region (RR)
• represents the region for natural sway (±1 degree)
• is processed using the subject’s height
ML acceleration ML acceleration
Volume Balance
Volume Frequency
AP accelerationAP acceleration
A B
C D
SRRR
SRRR
Example of ABF signals
• ABF can provide similar information as one otolith:– If the trunk/head moves slowly, primarily
gravitational information is provided– If the trunk/head moves quickly, primarily
acceleration information is provided
• Continuous ABF sound also provides trunk VELOCITY information (most critical)
Remarks
• Subjects learn how to use Diapason in 1 minute
• Subjective balance score (Schieppati et al., JNNP 1999) is lower also when ABF seems NOT actually helpful
• It is small, light and comfortable to wear
Release 2 - 2003
Results: quiet standing
• Improve balance (Sway Area decreases)• Increase control (Mean Velocity increases)
ABF can restore standing stability in patients
NO ABF WITH ABF
This subject can NOT stand on the foam with eyes closed.
This subject can stand on the foam with eyes closed using ABF.
0 50 [mm]
CO
NTR
OL
VEST
IBU
LAR
Eyes Closed Eyes Open and foam Eyes Closed and foam
AB
F ON
AB
F OFF
AB
F OFF
AB
F ON
NO ABF
ABF
NO ABF
ABF
Which is the origin of the changes observed in body sway
trajectories?
Cognitive
Sensory
Tuning-fork
??Open-Loop
Feedback
Feed-forward
Some insights on the structural properties of body sway can be
provided by Stabilogram Diffusion
Analysis (Collins & De Luca, 1993)
Stochastic analysis: fractional Brownian motion (fBm) modeling through the
Variance Analysis Method
We model the output of the human postural control system as a system of (1-D and 2-D) bounded, correlated random walks.
This is done by investigating the memory of the system through an analysis of the increments in displacement (x, y, r)
Vx(t) = <x2> - <x>2 t2Hx
Vy(t) = <y2> - <y>2 t2Hy
Vr(t) = <r2> - <r>2 t2Hr
C = 2 (22Hj-1 - 1)
For fBm the following correlation function holds (Feder, 1988):
Computed from the experimental data
Estimated by LS techniques
0.05 0.310
-1
100
101
HjS > 0.5j = x, y, r
C > 0 (persistence)
0.5 110
0
101
102
HjL < 0.5j = x, y, r
C < 0 (anti-persistence)
Hence, COP fluctuations have a structure that is dependent upon the timescale of observation and not simply random (fractality).
Moreover, AT LEAST TWO scaling laws are needed to accurately model the phenomenon in the range of interest (0.01- 10 s).
- Scaling functions that describe how the values change with the resolution tells more about the data than the value of the measurement at any one resolution (in particular at the higher resolution as it is commonly done by the summary statistic scores, working with the original sampled time series).
- Two modes of postural control take place over different timescales, associated with persistent and anti-persistent motion of the COP
Collins & De Luca, 1993
HOW MANY SCALING REGIMES ?This mechanism of transition from persistent to anti-persistent behavior is a common property of many biological systems. How to model this ?
We choose the easiest answer: correlation is assumed to change continuously
10-3
10-2
10-1
100
101
102
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
H
c
tKtV
tttH
2)(
)]/1(2log[2log)(
Chiari et al., Hum Mov Sci, 2000
H
t
-Parameter K is the variance of the displacements for t=1 s, which is also proportional to the variance of the displacements for large time-lags (V(t→4K as t→∞). Since V(tc)=K tc, K can be thought of as an estimate of the actual diffusion coefficient of the random process which is encountered by sampling the time series r at a sampling frequency 1/tc.
-In fact, parameter tc is the midpoint of the sigmoid and represents the time-lag in which H = 0.5, corresponding to a purely random behavior. In this sense it is an estimate of the time-lag at which the real process switches from a persistent (positively correlated) to an antipersistent (negatively correlated) behavior.
K & tc do correlate with several pathological conditions Central pathologies - Parkinson (Rocchi et al., 2000) - Multiple sclerosis (Chiari et al., in preparation)Peripheral pathologies - Peripheral Neuropathy (Lenzi et al., 1999) - Vestibular Loss (Kluzik et al., 2001)
K & tc allow to identify different postural strategies in control subjects - Post-adaptation to an incline (Chiari et al., 2001;) - Postural blindness (Chiari et al., 2000)
Do structural properties of the postural sway change with ABF?
If so, in which way this may help in the understanding of the
mechanisms underlying ABF efficacy and improving the design
of a rehabilitation strategy for balance disorders?
100.0
266.7
433.3
600.0
1 2
ABF_on_off
K
* **
0.1
0.2
0.3
0.5
1 2
ABF_on_off
delta
_Tc
*p<0.05**p<0.01
ResultsWe present the results obtained from 9 healthy subjects in the condition with the least sensory cues (i.e. eyes closed on foam) that benefited the most from ABF
Both K and tc show a systematic reduction due to ABF
Structural properties do change during biofeedback trials
Conclusions• ABF is comfortable and well accepted by the subjects• Subjects increase postural control using ABF (area
decreases, mean velocity increases)• ABF may help people manage more easily inadequate
surface somatosensory and visual information for postural control
• ABF determines structural changes in the COP that may reflect a larger role for feedback (conscious?) control over feed-forward control of posture.
• Future studies are needed to determine whether, with more practicing, subjects can use ABF without conscious control and hence how much this result is consistent over time.
Thank you for your attention
Luigi Galvani, Guglielmo Marconi and Augusto Righi, Bononiensi
Work in Progress
• Development of a portable wireless prosthesis for balance improvement
• Use in clinical rehabilitation for subjects with balance deficits
• Validation of ABF during dynamic tasks
Open question: Can use of ABF become more automatic with
practice?
• We have shown that practicing with ABF increases subject’s balance performance
• Vestibular loss subjects have difficulties using ABF when they are already controlling balance using a voluntary strategy i.e. concentrating specifically on the other senses (Divided Attention problem). Can use of ABF become more automatic (less voluntary)?