Background
Spinal cord injury (SCI) decreases overall quality of life due to a plethora of potential adverse effects to many body systems that compromise musculoskeletal, urinary, cardiovascular, and other essential functions. Quality of life is heavily reliant on the level of physical function, with individuals capable of a higher level of independence experiencing a higher quality of life [
1,
2]. Health preference scores are a measure of health-related quality of life and have been reported as 0.93 for the general population [
3], 0.58 for stroke survivors [
4], and 0.42 for those with Parkinson’s disease [
5]. Because SCI severely compromises motor and sensory function below the level of injury this population has one of the lowest health preference scores of any disability group at 0.27 [
3]. It is critical to improve the intrinsic function of individuals with SCI as even small increases in their physical capabilities will have drastic effects on quality of life and independence.
An under-examined aspect of functional independence is the role of trunk stability in facilitating activities of daily living (ADLs) by providing support against destabilizing perturbations while using the upper extremities. Trunk movement enables ADLs by enlarging the workspace [
6] and are even employed when an object or ADL is within arm’s length [
6,
7]. For these reasons, trunk stability has been consistently rated as a high priority for improvement after SCI [
8‐
10]. Trunk stability is also an important factor for preventing falls in individuals with SCI [
11] and can promote balance and stable function in the wheelchair [
12].
Functional neuromuscular stimulation (FNS) has successfully provided options for people with SCI to independently perform many activities affected by paralysis, including reaching [
13], biking [
14], and walking [
15]. Applying a constant level of stimulation to the nerves serving the hip and trunk muscles improves seated posture during quiet sitting [
16] and increases reaching distances [
17]. However, stimulation needs to be modulated appropriately to compensate for different loading scenarios and applied perturbations that could destabilize seated posture. Perturbations can arise from both external (e.g., wheelchair collisions) and internal (e.g., manipulating heavy objects) sources. Various forms of feedback controllers have been proposed to maintain seated balance in the sagittal plane by adjusting stimulation parameters to resist external perturbations applied to the trunk [
18,
19]. Murphy et al. [
20] and Bheemreddy et al. [
21] implemented a threshold-based self-righting controller to recruit trunk muscles with supramaximal stimulation once the subject exceeded a predetermined trunk flexion or bending threshold to effectively return the user to an erect posture from large forward flexed or sideways bending postures. Proportional, integral, derivative (PID) control was found to be particularly appropriate for upright sitting since the extrinsic (active muscles) properties of the trunk are dominated by the proportional and derivative model components [
22]. The goal of this study was to expand these previous control architectures to maintain stability in the presence of internal perturbations generated during both sagittal and coronal plane movements while performing functional reaching tasks simulating typical ADLs. The intended effect of the PID control law was to increase levels of stimulation whenever the trunk moved away from the baseline erect position, thus returning the subject to erect after the internal perturbations and minimizing the resultant overall postural sway.
The objectives of this study were to: (1) implement a multi-directional controller to maintain upright seated posture, and (2) determine the effects of the controller on posture during a functional reaching task. We hypothesized that the controller would reduce the postural sway during a functional reaching task and will quicken return to erect movements compared to without stimulation or applying low levels of stimulation to ensure a baseline constant stiffness of hips and trunk.
Discussion
Maintaining a stable base of support is necessary to complete many ADLs, as they often impart internal perturbations as a consequence of moving objects from one location to another. Upper extremity actions, including driving, eating, and dressing all impart unique destabilizing disturbances on the trunk, requiring an equal and opposite reaction to counteract the effect of the disturbances. We have demonstrated a feedback control system capable of maintaining an upright posture in response to reaching tasks by simultaneously controlling trunk extension and lateral bending in individuals with paralysis.
Ellipse area, while typically utilized to analyze posturography during quiet sitting [
31,
37], is applicable to understanding the range of trunk movement our subjects employed to accomplish the reaching task [
38]. We initially hypothesized that feedback control would reduce the ellipse area, as larger movements away from erect results in increased levels of stimulation based on the control logic. Subjects S1 and S4 both showed a significantly reduced ellipse area, supporting this hypothesis. A non-significant reduction was observed in S2 and S3 and a trending, but non-significant increase was observed for S5. The compensatory strategies employed by those subjects likely influenced this outcome. For example, while every subject was asked to place their non-active hand on their stomach and only employ it in the case they sensed an impending fall, the non-active arm was sometimes used to press the hand on the abdomen. This may have aided in stiffening the trunk by increasing intra-abdominal pressure which has been associated with increases in spinal stability [
39,
40], much like wearing an abdominal belt has been shown to increase lumbar stiffness [
41]. It is possible pressing on their abdomen achieved a similar result.
Three subjects (S1, S2, S3) saw significant increases in mean effective unilateral reach distance with feedback control. This indicates these subjects chose to or were able to hold the weighted jar farther away from their center of mass when the controller was active, which is consistent with the effects of constant levels of FNS applied to trunk muscles during bimanual reach [
16,
17]. In contrast, S4 experienced a significant decrease in mean effective reach as well as a decrease in ellipse area or overall trunk movement with the controller. During the feedback control condition, S4 immediately withdrew her arm to her chest after releasing the jar to the target stations, while during the no stimulation case she commonly placed her arm on her thigh to catch herself in instances of instability, resulting in a lower mean effective reach despite decreased trunk movements when the controller was active.
A major benefit of the feedback control is improved return to upright motion. The controller quickened the return to upright motion of four subjects during deployment of the jar to the target stations. Improvement ranged from 0.17 to 0.32s, a decrease of between 15 and 25%, compared to the no/low stimulation case. While the difference is small for a single reaching movement, implementation of feedback control at home can quicken return to upright every time the trunk is away from erect. Individuals with SCI typically seek to maintain an upright posture while reaching by employing strategies that provide counterbalance to the reaching arm such as extension of the trunk [
33]. Returning the subject to an upright stable posture faster after a reach will reduce the length of time in an unstable posture. In instances with possible falls it is also crucial to restore upright sitting as fast as possible [
21,
42]. The return to upright movements are likely a combination of eccentric and concentric contractions. Eccentric contractions of hip and trunk extensor muscles occur when the controller activates during deployment of the trunk, as those muscles would be lengthening. On the other hand, when the subject is returning to upright, the contractions of those muscles become concentric. This organization possibly benefits the return to upright motion as the stronger eccentric contractions serve to counter the initial momentum of the trunk during deployment, while the concentric contractions return the user upright.
Prior experiments have demonstrated the ability of a threshold-based controller to return a person to an upright position after a large trunk disturbance [
20,
21]. These benefits continued to be evident during smaller internal disturbances caused by an upright reaching task in the current study. Interestingly, the return to upright motion was minimally affected by stimulation when the subject returned the jar to the home station. One possible explanation is due to the subjects pushing down on the table with their upper extremity through the jar, thereby propelling their trunk to the upright position without need for stimulation assistance. This compensatory strategy has been reported before during bimanual reaching tasks [
33,
43] and was observed in all our subjects. It was especially prevalent in subject S5, who has a complete injury at the T10 level and as such retains more voluntary control of his trunk musculature, which may be why the controller had limited effects on his movement. This suggests that it may be reasonable to expect that a neuroprosthesis for seated reaching or postural balance will have greatest benefit to recipients with higher levels of injury and least control of their hip and trunk musculature. A similar reliance on injury level was also reported while assessing the effects of constant stimulation on hip and trunk muscles during various functional activities [
17].
Previous work on FNS of the trunk has focused on bimanual tasks [
16,
17], thus eliminating the majority of compensatory strategies available to individuals with complete SCI. Here we attempted to create a realistic unimanual task by having subjects in their own wheelchairs and allowing them to catch themselves in instances of instability. This resulted in compensatory strategies including increased hand pressure on the abdomen, pressing down on the jar to return the trunk to erect, and placing the active arm on their thigh to prevent forward trunk pitch. The full impact of a neuroprosthesis for trunk stability on reducing or eliminating reliance on such compensatory measures remains to be investigated. In other disability conditions where there is impaired control of the upper limbs, such as in stroke survivors [
44,
45] and upper limb prosthesis users [
46], exaggerated trunk movements are commonly utilized and required to perform reaching tasks. Facilitating trunk movements that complement the desired reaching motions would likely change or possibly reduce the compensatory strategies exhibited by the subjects in this study.
While a qualitative study, the participants did express a preference for reaching with the controller active. Perturbation resistance is important in many activities of daily living. Our subjects have expressed interest in utilizing improved trunk stability for everyday tasks, such as driving, returning to upright after acquiring items, and food preparation and cleanup. Improved seated stability has been correlated with improved completion of a variety of upper-body tasks, including dressing [
47], typing, eating, moving boxes, opening doors, etc. [
48]. Designing a controller that is both resistant to perturbations and capable of holding upright or leaning postures could improve seated stability thus aiding in these activities of daily living.
There are several limitations of this study. First, the controller was tested with five subjects and showed encouraging, though varied effects. The control system should be applied to a larger group of users with different presentations before the results can be generalized to the entire population. Second, trunk axial rotation (twist) was ignored and only trunk flexion and lateral bending were addressed. Currently, control of axial rotation was not achievable with the muscle groups being activated by our neuroprosthesis and would require targeting additional trunk muscles such as the internal and external obliques. Incorporating axial rotation may require attention in the future especially for reaching of objects directly to the side of the subject. Third, muscles were assigned to a given motion (flexion, extension or lateral bending) and stimulated based on the same control signal regardless of their specific recruitment properties. Future control schemes should seek to identify and incorporate subject-specific muscle synergies and recruitment properties. Fourth, the feedback signal for control was obtained from raw voltage signals of one accelerometer placed on the sternum. An improved measure of trunk position could be obtained with orientation data from a 9-axis inertial measurement unit (IMU) or a network of sensors combined to yield a high fidelity estimate of trunk orientation. Additionally, because the neuromusculoskeletal system is nonlinear, non-autonomous, and partially unobservable, the linear PID control law is only valid for a limited operating range of the trunk and pelvis [
49]. In the upright seated position, the proportional and derivative terms dominate the extrinsic trunk properties [
22], suggesting that a PID controller is valid during quiet sitting. Here we may have exceeded the linear operating range with trunk excursions of over 20° (Fig.
7b and c). This is a possible explanation of why the improvement in the return to erect motion was most evident in the closer ipsilateral station movements and less prevalent during reaching toward the midline and contralateral stations that required greater trunk excursion (Fig.
7). Future work should explore nonlinear control schemes, such as a model predictive controller [
50] or a linear controller in combination with knowledge of the subject specific nonlinear muscle properties. Finally, it is possible that despite best efforts, fatigue may have accumulated within testing sessions and impacted results. We minimized the effects by resting between each trial and ensuring that each subject completed recondition exercises at home with the system for several months before the experiments.
Future work should also focus on translating the controller out of the laboratory and into the user’s home. The Networked Neuroprosthesis is a promising target for integrating these controllers in everyday life due to containing the necessary electrodes, implanted accelerometers and processing capabilities [
51]. Through a home-going trial we can directly examine the effects of feedback controlled trunk movements on the user’s quality of life and the direct effects on activities of daily living.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.