Stroke leads to abnormal coordination of the multiple upper extremity joints during passive movement

Background Motor impairments in the upper extremity (UE) is one of the most common deficits after stroke. Even though understanding of UE coordination deficits in persons with strokes is critical for better identification of motor impairment and planning for rehabilitation, it is still not clear how stroke affects coordination patterns of multi-joint movements in the UE. Here, we investigated kinematic and kinetic coordination patterns of UE after stroke during controlled passive arm movement.


Background
Motor impairments in the upper extremity (UE) is one of the most common deficits after stroke. Even though understanding of UE coordination deficits in persons with strokes is critical for better identification of motor impairment and planning for rehabilitation, it is still not clear how stroke affects coordination patterns of multi-joint movements in the UE. Here, we investigated kinematic and kinetic coordination patterns of UE after stroke during controlled passive arm movement.

Methods
An exoskeleton multi-joint robot moved the participant's arm in the horizontal plane back and forth in 8 repetitions, in inward movement (i.e. toward the body) and outward movement (i.e. away from the body). The uncontrolled manifold analysis (UCM) was used to quantify kinematic and kinetic coordination patterns of the UE. Variability of joint angles and torques were decomposed into task-relevant variability (TRV) and task-irrelevant variability (TIV). An index of coordination (IC) was defined based on TRV and TIV.

Results
We found that the IC of joint torques in the stroke group significantly decreased during outward movement in comparison to that during inward movement, while IC of the control group showed no difference between the two movement directions. The decreased IC in the stroke group during outward movement was mainly due to the increased TRV of joint torques. In the further analysis of individual joint level, during outward movement, stroke group had a greater TRV of joint torques at all joints while during inward movement, stroke group had a lower TRV of joint torques at elbow joint.

Conclusions
Our results indicate that the stroke can cause the kinetic coordination deficits induced during a passive movement especially in outward movement. Our findings suggest that it is important to consider the passive kinetic coordination deficits to enhance post-stroke rehabilitation interventions.

Background
About 795,000 people suffer strokes every year in the United States and stroke is the third leading cause of death. Stroke survivors often live with long-term motor functional impairments in daily activities [1]. Motor impairment in upper extremity (UE) is the most common deficit after stroke, reported in more than 80% of acute and 40% of chronic stroke survivors [2]. Even though there are several researches conducted to improve motor functions in stroke survivors by relearning and developing compensatory mechanisms, their impaired UE function for activities of daily living is still regarded as one of most challenging sequelae of a strokes [3][4][5].
It has been reported that after neurological deficit such as stroke, coordination patterns of UE movement are altered [6][7][8]. Clinically, the term synergy has been used to describe coordination deficits evident in abnormal co-activation of muscles after stroke. Abnormal synergies after stroke involve a tight coupling, respectively [9]. Such abnormal couplings have been observed in previous studies of UE kinetics reporting abnormal torque patterns between elbow extension and shoulder adduction in isometric maximum voluntary torque [10,11].
However, these studies provide little information about the underlying mechanics to reveal how the central nervous system (CNS) generates abnormal coordination patterns of UE movements.
In recent years, the concept of synergy has emerged in neuroscience and motor control as a fundamental mechanism of the CNS to control multiple degree of freedom of the motor apparatus for desired inter-joint coordination. Motor synergy has been examined under the framework of Uncontrolled Manifold analysis [12,13]. The UCM analysis provides a quantitative approach of quantifying how multi-joint movements are organized or controlled by examining the structure of variability in the elemental variables (e.g. arm joint angles) with respect to task variables (e.g. hand position Cartesian coordinates). In the arm reaching task, for example, the UCM quantifies the variance across trials of joint combinations partitioned into two components at every point in the hand's trajectory; joint variance that led to a consistent hand position (i.e. TIV, task-irrelevant variability) and joint variance that led to an inconsistent hand position from trial to trial (i.e. TRV, task-relevant variability). The UCM approach has been applied to several voluntary motor tasks. In healthy populations, many studies found that TIV was greater than TRV, indicating more flexibility in utilizing degree of freedoms (DoF) to accomplish a task. In a person with neurological diseases such as Parkinson's disease [14], olivoponto-cerebellar atrophy [15], multiple sclerosis [16] and spinocerebellar degeneration [17], different patterns of motor synergies have been observed often with decreased TIV and increased TRV as compared to healthy groups. However, previous stroke studies have reported inconsistent results on TRV and TIV between stroke group and healthy group during a UE reaching task [18][19][20]. One of the possibilities for the inconsistent results may be attributed to changes in passive properties of UE after stroke. Several previous studies have found increased non-reflex passive joint stiffness in subjects with chronic stroke [21,22]. Such an increased stiffness may be reflected in formation of abnormal variability of UE.
Recently, advanced robotic technologies have made it possible to systematically investigate and assess multi-joint movements [23]. In our previous work [24], a multi-joint intelligent rehabilitation robot, named IntelliArm, was developed to control the shoulder, elbow and wrist individually and simultaneously. IntelliArm allows us to evaluate more quantitative and systematic characterizations of neuro-muscular changes across multiple joints. Here, using IntelliArm, we investigated how stroke affect UE coordination patterns during passive movement.
Hypothesis is a two-fold: 1) abnormal coordination patterns would be more severe during outward movement (i.e. away from the body) as compared to inward movement (i.e. toward the body) due to stroke accompanied by flexor hypertonia of UE and 2) wrist joint would contribute most to the abnormal coordination patterns, based on the classical perception of a proximal to distal gradient of motor deficits in UE after stroke (i.e., distal segment is more severe than proximal segment) [9,25].

Participants
10 stroke survivors and 10 age-matched control subjects were recruited for this study. All the control subjects were right-handed (8 males and 2 females, mean age 50 years, SD 12.4). Three stroke survivors were right hemiparetic, and the other seven patients were left hemiparetic (8 males and 2 females, mean age 62 years, SD 8.4). Inclusion criteria of the stroke patients were a stable medical condition, an interval of at least 1 year since stroke onset with cognitive ability to follow simple instructions. Prior to the experiment, each participant gave a written consent approved by the institutional review board of University of Maryland, Baltimore.

Experiments
In order to assess UE function, we used a rehabilitation robot for neurorehabilitation of the shoulder elbow, and wrist joints we developed previously (Fig. 1a). The robot called IntelliArm was designed to perform a two-dimensional motion on the horizontal plane with gravity support of UE. All subjects were asked to sit upright comfortably, and to put the upper arm, forearm and hand on IntelliArm (Fig. 1a). The robot arm lengths for the upper arm and forearm were then adjusted to align mechanical axes for shoulder horizontal adduction/abduction, and elbow and wrist flexion/extension with the corresponding subject's anatomical joint axes.
The subject's upper arm and forearm were strapped to the corresponding braces to ensure wellalignment throughout the experiment. Initial position was at shoulder horizontal adduction of 70°, elbow flexion of 60° and wrist flexion of 0°, respectively (Fig. 1b).
IntelliArm produces simultaneous movements at shoulder, elbow, and wrist joints, starting at initial position to outward direction followed by inward direction. Each joint was moved until it reaches pre-specified limits for either the joint resistance torque or angular position [26]. Torque and angular position limits were set to be ±3 N·m and 0° to 120° for shoulder horizontal adduction/abduction, ±2 N·m and 10° to 90° for elbow flexion/extension, and ±1.5 N·m and -45° to 45° for wrist flexion/extension, respectively. There was 3 seconds holding time at initial position and all movements were performed for 8 repetitions (Fig. 2). Joint torques and angles were recorded simultaneously using a torque sensor and motor encoder, respectively. Each joint was moved at slow speed of 10°/s was selected in order to minimize reflex mediated actions and manifest passive mechanical properties [22].

Uncontrolled manifold analysis
The UCM analysis has been used to assess multi-joint coordination patterns in a redundant system by quantifying variability of elemental variables (i.e., joint angles or torques) with respect to changes in performance variable (i.e. position or force or the tip of the endeffector) [27]. Using the UCM analysis, we examined how multi-joints interact each other to stabilize the actions of the end-effector in terms of kinematics and kinetics.

Kinematics
In order to quantify coordination patterns of multi-joint angles, the 2-D position of the tip of the end-effector (i.e., the tip of the index finger or the tip of the rod) was selected as performance variable, whereas joint-angles were selected as elemental variables. First, a forward-kinematics model in horizontal plane movement was created for the human arm as follows: The null space of J represents the changes in elemental variables that do not lead to a change in the performance variable, referred to as task-irrelevant space, whereas the orthogonal component of J represents the changes in elemental variables that do lead to a change in the performance variable, referred to as task-relevant space. The basis vectors, ε, for the null space was calculated using Matlab null() function such that Jε = 0.
Two types of variability, the deviations from the average trajectories in joint space projected onto task-irrelevant space, , and task-relevant space, , respectively were computed as follows.
We calculated variance of and to measure TIV and TRV of joint angles, respectively.
where is a number of trails and here n = 8.
Note that and were transformed to correct for a non-normal distribution using the log transformation [28].

Kinetics
Similar to quantification of kinematic coordination, we quantified kinetic coordination patterns of UE. Using the Jacobian matrix calculated, a linearized task equation, force of the endeffector as a function of joint torques was computed.
The basis vectors, ε, for the null space of + was calculated such that + ε = 0.
The variabilities, the deviations from the average trajectories in joint space are projected onto task-irrelevant space, , and task-relevant space, , respectively.
The variance per joint of the projected deviations is where and are dimensions of task-irrelevant space, = − , and task-relevant space, = − , respectively. and are a number of joints (=3) and a dimension of a task space (=2), respectively.
Note that and were transformed to correct for a non-normal distribution using the log transformation [28].
An index of coordination (IC), the ratio between these two variances, quantifies to what extent the joint variables are coordinated to keep the target trajectory in task space.

= − +
Importantly, this ratio is not necessarily related to the overall joint-space variance; instead, it quantifies its structure with respect to the task space. So far, during an active movement, has been used to investigate the CNS's control mechanisms about the extent of stabilization of the particular task variable. When is larger than zero, a higher IC indicates that joint space variables are more coordinated such that the task variable is stabilized. If is equal to or smaller than zero, the coordination in joint space is indifferent to the particular task variable, or even destabilizes it.

Statistical analysis
A three-way repeated measures ANOVA with factors Joint (3 levels: shoulder, elbow, and wrist), Direction (2 levels: inward and outward) and Group (2 levels: stroke and control) were used to test the differences between conditions. In addition to this, a two-way repeated measures ANOVA with factors Direction and Group was performed to test differences of index of coordination. The level of statistical significance was set at p = 0.05. A post-hoc test was performed where necessary.

Index of Coordination
IC of joint angles, and during inward and outward movements were depicted in Figure 3. In both and , there was a significant main effect of

Discussion
The aim of this study was to investigate how stroke affect UE coordination patterns during passive movement. Using the UCM approach, UE coordination patterns were analyzed and quantified as IC, a ratio of TRV and TIV. We expected that abnormal coordination patterns would be more severe during outward movement as compared to inward movement due to stroke with flexor hypertonia. Kinematically, we found no significant coordination patterns between Stroke group and Control group. However, kinetically, we found that index of coordination, IC, in Stroke group significantly decreased during outward as compared to during inward while IC in Control group remained unchanged between movement directions. Our finding of the decreased IC in Stroke group during outward movement was mainly due to the increased TRV of joint torques. These results indicate that during passive stretching, joint torques were generated in a way that force at the end-effector (i.e. hand) became more variable or less inconsistent during outward movement than inward movement. In other words, Stroke group had abnormal coordination patterns of joint torques during outward movements by producing joint torques that resulted in inconsistent generation of force at the hand.
Our results did not support another hypothesis that the wrist joint would contribute most to the abnormal coordination patterns based on the classical perception of a proximal to distal gradient of motor deficits in UE after stroke. The classical perception of clinicians about individuals who have recently suffered a stroke is that paresis of distal upper limb segments is more severe than paresis of proximal upper limb segments [9,25]. Based on this clinical perception, it is often presumed that the loss of distal segment movement is responsible for the loss of hand function after stroke [29]. However, other studies found contradictory findings, reporting that a proximal to distal gradient of motor deficits is not present post stroke [30]. Our results in the current study in consistent with these studies suggested proximal to distal gradient of motor deficits in UE after study would not exist.
Our observations extend our understanding of stroke in the context of coordination patterns of UE movements. In the current study, it was the first attempt to use the UCM approach for coordination patterns during involuntary movements or passive stretching. So far, the UCM approach has been successful used to investigate the CNS control mechanism underlying coordinated movements during several voluntary tasks [12,[31][32][33][34][35]. The idea behind motor coordination in the framework of the UCM approach is that the CNS utilizes element variables in a way to ensure stability of performance variable, which reflects in task-specific pattern of distribution of the element variables, referred as to the principle of abundance [36]. In support of the principle of abundance, several studies suggested that have a greater values of TIV over TRV would be beneficial in dealing with unexpected perturbations [37], fatigue [38], and secondary tasks [39]. In pathological and gerontological studies [14][15][16][17], abnormal coordination patterns have been observed indicating pathological and gerontological markers of deficits or changes in the CNS control mechanisms for a redundant motor system. In the current study, however, considering that no or minimal active control by the CNS is involved during passive stretching, either positive, negative or zero of IC would not indicate good or bad (stabilizing or destabilizing) coordination on the CNS's control, rather it represents passive coupling patterns of inter-joint.
Any differences on IC in stroke group as compared to control group may indicate that passive structure of abnormal coordination patterns. Our findings that stroke group had abnormal kinetic coordination across the joints, possibly tied to the increased coupling in spastic upper extremity [21,22]. Thus, our findings of decreased IC in Stroke group during outward movements implies that increased hyperexcitability of the stretch reflex after stroke may play a role in formation of abnormal coordination patterns especially during UE outward movements.
We found that stroke group had a significant higher both on TIV and TRV as compared to control group. Our results provide meaningful insights into understating of coordination deficits after stroke. In previous stroke studies, coordination patterns during voluntary movements were not consistent, reporting unchanged IC [18,40], decreased IC [41], and increased IC [20], along with unchanged TIV [18,40], and greater TIV [19,20]. However, TRV was found to be greater than in control [19,20,41]. The previous studies suggested that this unexpected finding might have resulted from large inherent variability of neural signals between motor cortex and effectors [42] or being more dependency of effectors in the patients with stroke [43]. However, this finding is relatively contradictory to the principle of motor abundance because this finding may imply that stroke survivors would have superior control abilities to utilize DoFs for completion of particular tasks. In addition, based on the finding of the current study, the inconsistent findings of IC in previous studies may be attributed to changes in passive coordination patterns. Thus, the findings of the current study can provide new aspects of coordination deficits and be used for better identification of abnormal coordination patterns during active movements by subtracting the passive components of abnormal coordination patterns, which warrants further investigations.

Conclusions
In summary, our findings clearly demonstrate that passive UE movements in stroke individuals induces abnormal coordination patterns especially in outward direction. Such an abnormal coordination patterns in stroke survivors was mainly due to increased variability of joint torques that led to inconsistent end-effector (or hand) force compared with healthy adults.
Therefore, the abnormal coordination patterns provide important information for understanding of the CNS's control deficits following stroke.

Authors' contributions
Author contributions: K.K., DW. K., and LQ.Z., designed research; K.K., DW. K., and C.Z., performed the experiments; K.K., DW.K., G.O., C.Z., and LQ.Z., analysed data; K.K., DW.K., G.O., C.Z., G.K., and LQ.Z., wrote the paper. Figure 1. A multi-joint intelligent rehabilitation robot, named IntelliArm developed to control the shoulder, elbow and wrist joints simultaneously (a). The shoulder horizontal abduction/adduction, and elbow and wrist flexion/extension are controlled by the shoulder, elbow and wrist motor, respectively. The robotic arm is mounted on a large X-Y table allowing the IntelliArm robot to well-align the mechanical shoulder axis with the horizontal rotation axis of the subject. Schematic diagram that the subject is seated with the upper arm, forearm and hand attached to the IntelliArm (top view in horizontal plane) (b). IntelliArm produces simultaneous movements at shoulder, elbow, and wrist joints, starting to outward direction followed by inward direction. Note that shoulder horizontal adduction, elbow flexion and wrist flexion are positive. For analysis of upper limb coordination, a task equation of end-effector position ( , ) and force ( , ) was expressed as a function of joint angles and torques, respectively.   There was a significant interaction effect Direction × Group on IC. The following pair-wise comparisons revealed that IC in stroke group was a significantly lower during outward direction than inward direction while IC in control group remained unchanged between directions. The decreased IC in Stroke group was main attributed to the increased TRV during outward movement as compared to inward movement. The asterisk indicates a significant difference (*<0.05) between directions in stroke group. Error bars represent SEM across subjects.  ). During inward movement, TRV at elbow joint in Stroke group was significantly greater than in Control group. However, during outward movement, TRV at all joints in Stroke group were significantly greater than in Control group. During both directions, TIV at shoulder joint in Stroke group was significantly greater than in Control group. The asterisk indicates a significant difference (*<0.05; **<0.001) between groups. Error bars represent SEM across subjects.