Procedure
Each participant seated on a chair and gripped the handle of a planar manipulandum, the Inmotion2 Robot (Interactive Motion Technologies Inc., Boston, MA, USA), used to guide and perturb movements during the experiment. Trunk movements were prevented by means of a belt, while the elbow was supported in the horizontal plane by an anatomical orthosis. Subjects were instructed to move from the centre of the workspace forward and backward to reach eight different targets positioned every 45° on the perimeter of a circle with a 14 cm diameter. Subjects performed pointing exercises both in null force field (NF) and in a velocity-dependent force field (VF):
(1)
where forces were always orthogonal to hand velocity, forming a clockwise curl field (λ = 20 N s/m, v = hand speed). Such experimental paradigm has been used in several studies on motor control adaptation in altered force fields environments [
4,
18,
19].
Each subject involved in the study performed a total of 832 movements corresponding to 52 turns, divided into the following experimental session:
Session 1: Null field environment
exercise 1: Familiarization (2 turns to take confidence with the robotic device)
exercise 2: Learning unperturbed dynamics (20 turns in NF to learn how to move in this condition)
Session 2: Velocity dependent force field environment
exercise 3: Early learning (4 turns in VF field)
exercise 4: Adaptation (20 turns in VF field)
Session 3: Null field environment
exercise 5: De-Adaptation (4 turns in NF field)
exercise 6: Final Washout (2 turns in NF field).
Two further elderly subjects (group 1.2, 70 and 81 years old) executed the same protocol doubling the number of trials in exercise 5 of session 3 (de-adaptation phase). This approach was used to verify whether difference between the two groups at the end of the experiment could be related to fatigue or other physical factors.
Participants were instructed to perform movements in the most ecological way. During the experiment an audio feedback was given when they went too slow or too fast so that movement speed remained always between 0.15 m/s and 0.4 m/s. The purpose of this approach was to make them execute the exercise in the most natural way, in order to observe the real strategy adopted during the adaptation, but trying to obtain comparable performance inside each group. Visual feedback of target position while performing the exercises was given by a computer screen located in front of the subject. No explicit instructions regarding the hand path were given. Movements were recorded with the use of an Optotrak 3D optoelectronic camera system (Optotrak 3020, Northern Digital, Waterloo, Ontario Canada), and collected considering each trial as the displacement from the center to the goal point and back at 200 Hz sampling rate. The infrared diodes were positioned in four anatomical landmarks: trunk (sternum), shoulder (acromio), elbow, and wrist (considered as the end point).
Data analysis
Data were low-pass filtered (fifth order Butterworth filter, zero-phase distortion; MATLAB "butter" and "filtfilt" functions). Hand position was differentiated to compute speed, acceleration and Jerk profiles. Movement onset and offset were detected when the end-point velocity exceeded 5% of the peak velocity value. Shoulder and elbow joint angular displacements, velocities and accelerations were also determined. Positive direction of motion was assigned to flexion and negative to extension. Both kinetic and kinematic analyses were carried out by looking in a specific way at the different movement directions. In fact, other research groups [
20] have shown that the anisotropy and orientation of inertia ellipse of the upper limb determines movements characterized by higher inertia in left diagonal direction, and by higher accelerations in right diagonal direction. To evaluate the efficiency of movements a normalized length path parameter was calculated with the following Equation [
21]:
LL = (ΣdR)/L
t
where dR is the distance between two points of the subject's path and Lt is the theoretical path length, represented by the distance of the two extreme points of the stroke. Higher values of LL correspond to hand trajectories affected by larger errors.
The smoothness parameter N.Jerk was also computed using the metric proposed by Teulings and coworkers which consists of the time- integrated squared jerk opportunely normalized [
22]:
(3)
where j is the Jerk, that is the change of the acceleration per time, and it is calculated as the third time derivative of position. This parameter has the advantage to be dimensionless and usable to compare movements with different characteristics (i.e., duration, size). Reduced coordination results in multiple acceleration peaks at the base of an increase of the jerk levels, hence, the lower the parameter, the smoother the motion.
For each group, and for each movement direction the mean value and standard deviation of the movement smoothness have been computed within all the exercise sessions; in exercise 2 and 4 only the values of the last 5 trials were used in order to evaluate the values achieved after the consolidation of the learning process.
A simplified model of the arm based on the Newton-Euler [
23] recursive algorithm, was used to compute the torque acting at the shoulder and the elbow. Anthropometric measure of limb were took into account in the computation of the joint torques: segmental masses, location of mass centre and moments of inertia were estimated from he weight and the height of the subjects in accordance with Winter [
24]. Torques estimated at each joint with this model were grouped according to the approach proposed by Dounskaia et al. [
14]: 1) net torque (NT), proportional to the angular acceleration at the joint; 2) interaction torque (IT), that depends on motion at both joint and on the nature of the force field in which subjects moved; 3) muscle torque (MUSC) which considers the muscle activity and the viscoelastic properties of the entire arm. In particular, the Equations for torque computation at the joints are:
MUSE
E
= NT
E
- IT
E
- IT
field
MUSE
S
= NT
S
- IT
S
- MUSC
E
where
S and
E apexes represent the shoulder and elbow joints; IT
field = 0 when the field is turned off. To investigate the role of the MUSC, IT and IT
field components in motion production, a sign analysis was computed in accordance with previous works by Dounskaia and co workers [
14,
25]. Shortly, the torque sign analysis determines the percentage of time when the analyzed torque (MUSC or IT) has the same sign of the NT torque, i.e., it gives a positive contribution to movement acceleration and it is responsible for it. To provide information about the magnitude of the contribution of MUSC to the NET, the difference between positive and negative peaks of the MUSC torque was computed for both joints hence after called MT magnitude. The evolution of all these parameters (LL, N.Jerk, elbow and shoulder torques sign, and magnitude values) was monitored throughout the experiment in order to observe the macroscopic effects of different motor control strategies adopted by each person and group. Performance achieved by each subject at the end of exercise 2 were considered as a reference, i.e. subjects after being trained for a prolonged time in an unperturbed environment achieved the most ecological motion. Indeed, differences in kinematic and kinetic trends between exercise 2 and all the other phases were considered as a consequence of the presence of the external perturbation; their evolution during adaptation and de-adaptation was, then, used to quantify efficiency of the motor strategies adopted.