3.2 Control strategies for robot-based therapies
3.2.3 High-level: biofeedback strategy for postural control
Children with CP present an altered gait pattern with an increased ROM of the trunk during gait. Usually, they walk looking at the ground, with their head down. Some references in the bibliography ensure that to maintain a proper posture during walking
Figure 3.14: Right hip interval of 70s selected from the user’s performance with CPWalker using the multi-joint adaptive impedance controller. (a) represents the com- parison between the percentage achieved by the user each 3 steps (pink line) respect
to the desired challenge % (blue line). (b) shows the change of new (pink line) each
3 steps (blue line) depending on the percentage in (a). (c) is the graph of the user’s performance: yellow line is the real motion and blue line is the set point.
Table 3.1: Data for all the joints during the period of 70 s selected from the exercise with the healthy user. G0 to G6 are the groups of 3 steps implemented during this
period. The table gives the progression of new depending on the % achieved in the
last group of steps. Bold values represent % achieved> % , and (*) means % achieved
> 100% because the user’s flexion-extension set was bigger than that given by the set point.
G0 G1 G2 G3 G4 G5 G6
Right hip new P HI P HI MI HI MI
achieved 95.43 82.05 92.18 86.14 79.08 86.29 76.38
Left hip new MI HI P HI P HI P
achieved 78.64 77.11 91.17 82.19 97.09 84.59 95.67
Right knee new P HI MI LI LI LI LI
achieved 113* 100.6* 95.89 100.9* 95.27 102.7* 93.11
Left knee new P HI MI LI LI LI LI
is a relevant aspect, specially in the case of children with CP [88, 96, 97, 159]. These problems must be attended as independent movements limitations, and rehabilitation strategies must be oriented to correct them [160].
To address these issues, author developed an active postural control strategy based on CPWalker, which is intended to be used while over-ground movement in real environment is allowed. The rationale of this control strategy was to enhance the cognitive interaction between the child and the robot, improving the postural control through that.
IMUs sensors (TechMCS, Technaid, Spain) were used as the main part of a biofeedback strategy developed to improve the children’s postural control during robot-based gait therapies. Two IMU sensors, placed on the user’s chest and head, measure in real- time the orientation of the trunk and head respectively. The procedure based on this approach consists in giving acoustic feedback to the patients when they lose the control of a desirable orientation of the body. The range for a proper posture could be defined by the clinician at the beginning of the exercise, and the acoustic feedback, which is normally selected from disturbing sounds or alarms, alerts the child of the incorrect position. With this method, while the exoskeleton corrects the patients’ gait, the postural control strategy motivates the children to maintain a proper posture during ambulation. This information provided by the biofeedback strategy was a request of our clinical partners since it is a parameter of paramount importance during the execution of the robotic therapy [96,97,159]. The exercises with IMUs supported the correction of the patient’s crouched gait in order to achieve a better extended hip position, besides correcting the posture and improve motor control.
To monitor the orientation of trunk and head for the postural control strategy of CP- Walker, the information of both IMUs sensors (head and trunk) was sent to the clinician interface through rotation matrices (R, Equation 3.10):
R = 2 6 6 4 Xx Xy Xz Yx Yy Yz Zx Zy Zz 3 7 7 5 (3.10)
Each IMU sensor is referenced to the magnetic north and the maximum value of accelera- tion corresponds with z axis. This means that in order to measure rotations movements respect to an initial position, it is necessary to complete a calibration process [161]. During the execution of the therapy with CPWalker, the IMUs-based algorithm distin- guishes between the rotation matrix collected at the time of calibration (RG) and the rotation matrix captured by the IMU sensor in each instant (RS). With these data, a
new matrix (RGS) for the reference system is calculated as Equation3.11 indicates:
RGS= RS· (RG) 1 (3.11)
The algorithm obtains the Euler angles (↵, and , related with rotation in frontal, sagittal and transversal planes respectively) using the Equations 3.12, 3.13 and 3.14
based on the development carried out in [161]:
↵ = atan ✓ RGS(2, 3) RGS(3, 3) ◆ (3.12) = asin (RGS(1, 3)) (3.13) = atan ✓ RGS(1, 2) RGS(1, 1) ◆ (3.14)
In rest position, RGSis the identity matrix and consequently, the Euler angles are equals to zero.
The information provided by the sensors is represented in real-time after undergoing the conversion algorithms. Figure 3.15-right shows an example of recording data in real-time from IMUs-based system, where the measured angles for head and trunk (blue lines) in three spatial planes were compared to the ROM of the left hip during walking (red lines). Red squares in Figure3.15-left represent posture out of the permitted range (acoustic feedback playing).
3.2.3.1 Technical evaluation of the biofeedback strategy for postural control
The postural control therapy was preliminarily evaluated in one child with spastic diple- gia in order to assess the usability of the system in clinical practice. The exercise consisted on using the biofeedback strategy for postural control at the same time than the user was walking following position control in the exoskeleton of CPWalker. The training lasted 5 sessions of 40 minutes each, one session/day. The main objective of this trial was oriented to assess the motor control improvements of the trunk during gait.
With the aim of objectively measuring the progress of the subject after this robot-based therapy, trunk kinematic data was obtained from 3D gait analysis before and after the experiment. The data collection was performed using an eight infrared cameras system (BTS BioEngineering, Italy). Reflective markers were applied on the shoulder girdle
Figure 3.15: IMUs based interface to give biofeedback of postural control in head and trunk. The graphics show IMUs data collected in real time for head and trunk in three planes (blue lines), and these were compared with hip ROM (red lines). The red
squares represent postures out of the limit values (acoustic feedback playing).
(spinous process of C7 and both acromio-clavicular joints). Marker trajectories were processed and analysed. For comparisons, a pre-post graph was performed for the child (Figure 3.16). In this graph it is possible to see that post-training data (continuous lines) are closer to normal values (grey zone) than data from the pre-study (dotted lines). These preliminary outcomes reveal the potential of recovery of this strategy.
Figure 3.16: Patient’s trunk kinematics during the pilot technical experiment. Nor- mal trunk kinematics data is represented in grey. Pre-intervention data is represented through dotted lines. Post-intervention data is represented through continuous lines.
Figure 3.17: Schematic view of the methodology used for the CPWalker graphic interface.