Abstract
Recent studies have shown that supplemental sensory feedback systems have potential to mitigate functional impairment after neuromotor injury through mechanisms of sensory augmentation or replacement. In this proof-of-concept case series, we evaluated multi-session 3-dimensional kinesthetic vibrotactile feedback training as a means to enhance the accuracy and efficiency of goal-directed reaching in the absence of visual feedback in survivors of stroke. A motion capture system converted real-time position of the contralesional hand within a Cartesian frame of reference into spatiotemporal patterns of vibrotactile feedback provided to the non-moving, ipsilesional arm. Seven survivors of stroke underwent 9 hours of reach-to-grasp training under conditions that encouraged them to learn a mapping from hand position to patterns of vibrotactile feedback. We then assessed their ability to use that feedback to improve the accuracy and efficiency of reaches performed without concurrent visual feedback. Within-subject comparisons to baseline performance revealed heterogeneous learning effects on reaching both with and without supplemental vibrotactile feedback. When reaching with the supplemental feedback after training, three participants significantly improved reach accuracy, two significantly improved temporal efficiency, and three significantly improved spatial efficiency (although one significantly worsened with regard to spatial efficiency). Further, all but one of the participants who exhibited post-training performance changes during movements made with the supplemental feedback also exhibited similar post-training performance changes during movements made without the supplemental feedback. These results suggest that while a subset of stroke survivors may accrue benefits from using the form of supplemental vibrotactile kinesthetic feedback described in the paper, the effects of such training resulted primarily from ancillary benefits provided by the 9 hours of intensive reach-to-grasp training with the feedback, rather than from the ongoing use of the feedback itself. Despite the mixed kinematic results, responses to standard surveys of subjective experience suggest that training with the vibrotactile display had acceptable usability and was both motivating and satisfying to use. We conclude that while the wearable technology may provide a positive user experience, practical benefits (e.g., increased reach accuracy or efficiency) may accrue more from hours of focused reach training than from the additional sensory information provided by the vibrotactile interface. (National Clinical Trial Number: NCT03298243; clinicaltrials.gov/study/NCT03298243)
Keywords
Introduction
Decreased sensations of limb position and movement are experienced by ∼50% of stroke survivors, leading to a decrease in upper extremity motor control and difficulty performing common daily tasks (Rand, 2018). Although visual feedback of the arm and hand can partly compensate for deficits of kinesthetic sensation, resulting movements are typically slow and jerky (Sarlegna et al., 2006). Using supplemental vibrotactile feedback to enhance closed-loop control of paretic limbs has been proposed as a compensatory approach to addressing sensation deficits (Ballardini et al., 2021; Jouybari et al., 2024; Tzorakoleftherakis et al., 2015, 2016). One study from Dupin and colleagues (2015) showed that people can successfully combine tactile information provided to a nonmoving limb with the proprioceptive information from a moving limb to form a single coherent percept of limb motion. Building on that foundation, recent work has shown that healthy individuals and some survivors of stroke can use a form of contralateral sensorimotor integration of vibrotactile feedback about hand position to enhance the accuracy of 1- and 2-dimensional reaching tasks performed without concurrent visual feedback (Ballardini et al., 2021; Krueger et al., 2017; Rayes et al., 2023; Risi et al., 2019; Shah et al., 2023).
Experimental studies of supplemental feedback systems also can provide insight into fundamental mechanisms of multisensory motor learning. For example, providing neurologically-intact human subjects with supplemental vibrotactile feedback of hand position within peripersonal space has been found to substantially reduce the phenomena of “proprioceptive drift” and distortions of the reachable workspace size that are commonly observed when view of the arm and hand is occluded during reaching (Findlater et al., 2018; Findlater & Dukelow, 2017; Moore et al., 2022; Risi et al., 2019; Scheidt et al., 2005; Wann & Ibrahim, 1992). The fact that it takes only ∼45 minutes of training with the novel supplemental feedback to effectively eliminate drift and other proprioceptive perceptual distortions demonstrates in dramatic fashion the great extent to which the human sensorimotor control system flexibly combines information about kinematic performance across multiple sensory sources – even favoring a completely novel source over one with an accumulated lifetime of experience. More recently, a 20-day longitudinal study involving healthy young adults demonstrated that stimulating a non-moving arm with supplemental vibrotactile kinesthetic feedback of the moving arm's hand location in peripersonal space led to significant improvements in reach accuracy and target capture times that plateaued after approximately five hours of training, which was distributed over just 10 out of the 20 days of training (Shah et al., 2023).
Other studies have compared how different ways of encoding supplemental vibrotactile feedback impact its immediate utility with respect to enhancing the accuracy and efficiency of movements requiring just one or two degrees of freedom of motion. Potential encoding schemes explored include: continuous feedback of limb position and/or velocity relative to some reference body configuration [i.e., limb state encoding; (Ballardini et al., 2021; Krueger et al., 2017; Rayes et al., 2023; Risi et al., 2019; Shah et al., 2023)], continuous error feedback of the hand's location relative to a goal [i.e., target error encoding; (Ballardini et al., 2021; Bark et al., 2015; Cuppone et al., 2016; Kapur et al., 2009; Krueger et al., 2017; Lee et al., 2011; Lieberman & Breazeal, 2007)], optimal feedback cues relative to some objective cost function (Tzorakoleftherakis et al., 2016), and intermittent alarms indicating undesirable conditions (Ferris & Sarter, 2011). Comparisons of limb state and target error encoding found that (i) both encoding schemes can enhance stabilizing and reaching actions performed without ongoing vision of the arm and hand (Krueger et al., 2017), (ii) limb state encoding yields better results when hand position is primarily considered over reach velocity (Krueger et al., 2017), and (iii) target error encoding yields movements with somewhat greater accuracy than limb state encoding (Ballardini et al., 2021; Krueger et al., 2017). However, target error encoding poses a significant computational challenge because it requires the supplemental feedback system to infer, in real-time, not only the current location of the hand but also the location of the user's object of desire (i.e., goal-aware encoding). By comparison, limb state encoding does not need to be goal aware, effectively circumventing this challenge by offloading the estimation of hand positioning errors to the user. Doing so, however, increases cognitive demand during system use. To effectively use limb state encoding of supplemental kinesthetic feedback, the user needs to develop a mental map between the spatial geometry of the physical workspace and a representation of the hand's position encoded within the vibrotactile signals provided by the supplemental feedback system.
Most studies to-date have focused on movements performed in 1- or 2-dimensions (Ballardini et al., 2021; Bark et al., 2015; Krueger et al., 2017; Rayes et al., 2023; Risi et al., 2019; Shah et al., 2023; Tzorakoleftherakis et al., 2015); by contrast, functional interactions with real-world objects typically require movements in 3-dimensions. Few studies have considered providing supplemental vibrotactile feedback to enhance 3-dimensional movements in either healthy or stroke populations (see Bark et al., 2015; Mazorow et al., 2024). To address this gap, we performed a longitudinal training study wherein a small sample of stroke survivors performed reach-to-grasp movements using 3-dimensional supplemental vibrotactile kinesthetic feedback with limb state encoding of hand position. Based on previous pilot work showing that healthy young adults can demonstrate significant improvements in reach accuracy after only ∼2 hours of training with the feedback system (Mazorow et al., 2024), we sought to test two hypotheses: 1) survivors of stroke will also demonstrate improved accuracy and efficiency of reach-to-grasp actions when provided extended training with 3-dimensional vibrotactile feedback in the absence of concurrent vision of the arm and hand; and 2) survivors will find training with the supplemental vibrotactile feedback to be useful, motivating, and satisfying.
Materials and Methods
Participant Recruitment
A convenience sample of seven hemiparetic stroke survivors gave written informed consent to participate in this study [5 female; 63.3 ± 21.8 years old (mean ± 1 standard deviation); 5.1 ± 2.7 years since stroke] (Table 1). Exclusion criteria included: history of a bleeding disorder; fixed contracture, history of tendon transfer, or profound atrophy in the involved limb; diagnosis of any disease that might interfere with neuromuscular function (e.g., myasthenia gravis, amyotrophic lateral sclerosis); current use of medication that may interfere with neuromuscular function (e.g., aminoglycoside antibiotics, curare-like agents); history of seizures; psychiatric co-morbidities (e.g., schizophrenia); and concurrent illness or injury limiting the capacity to conform to study requirements. Participants went through a battery of clinical evaluations to determine eligibility per the inclusion criteria discussed in greater detail in the following paragraph. All procedures were approved by the Institutional Review Board of Marquette University (protocol: HR-3303) in accordance with the Declaration of Helsinki.
Demographic and Clinical Data for the Participating Subjects.
The Upper Extremity Portion of the Fugl-Meyer Assessment (FMA-UE) is scored [0,66] for the motor section and [0,12] for the sensation section. The Mini-Mental State Examination (MMSE) is scored [0,30]. †: mild impairment; *: moderate-to-severe impairment.
Clinical Evaluations and Participant Inclusion
A licensed physical therapist and certified exercise physiologist evaluated the status of sensory and motor function for each participant using a battery of sensorimotor assessments to determine the participant's eligibility to participate in this study. Additional cognitive testing was performed by a trained psychometrician. Altogether, the clinical assessments included Upper Extremity portion of the Fugl-Meyer Assessment [FMA-UE; (Fugl-Meyer et al., 1975)] to test motor impairment (score range: 0 to 66) and somatosensation in the contralesional arm (score range: 0 to 12); a tuning fork assessment of vibrotactile sensation in both arms across multiple dermatomes; two-point orientation discrimination to further assess tactile mechanoreceptor sense across multiple dermatomes (Tong et al., 2013); and the 30-question Mini-Mental State Examination [MMSE; (Folstein et al., 1975)] to assess cognitive functioning (score range: 0 to 30). Participant inclusion criteria included: diagnosis of at least one stroke; in the sub-acute or chronic stages of recovery (i.e., more than three months post-stroke); retained ability to give informed consent and follow two-stage instructions; at least mild impairment as assessed using FMA-UE (i.e., a score within the range [15,60]); and preserved tactile sensation in the ipsilesional arm determined by the tuning fork assessment.
Higher FMA-UE and MMSE scores indicate less impairment. Impairment levels were defined as follows: mild motor impairment: FMA-UE motor score [46, 55]; moderate-to-severe motor impairment: FMA-UE motor score [0, 45]; mild sensation impairment: FMA-UE sensation score [9, 10]; moderate-to-severe sensation impairment: FMA-UE sensation score [0, 8]; mild cognitive impairment: MMSE score [18, 23]; moderate-to-severe cognitive impairment: MMSE score [0, 17]. Table 1 shows participant demographics and clinical assessment scores.
Experimental Set-Up
Participants were seated in a high-backed chair that supported the contralesional arm against gravity (Figure 1A) [see (Housman et al., 2007, 2009)]. Participants held a squeeze ball that included both an integrated pressure sensor to monitor grip pressure and active infrared markers to monitor hand position. Participants wore a brace that restricted wrist motions without limiting their ability to squeeze the pressure ball. They additionally wore custom goggles that blocked the lower half of their visual field, thereby obstructing visual feedback of the moving arm (Figure 1A). A precision motion tracking system (Optotrak 3020; Northern Digital Inc., Waterloo, Ontario, CA) collected the 3-dimensional location of the squeeze ball as well as squeeze ball pressure in real-time at 200 samples per second. Hand positions (squeeze ball locations) were converted into cursor motions within a 56 cm×56 cm visual field displayed on a 50” (127 cm) vertical display (Insignia, NS-50F301NA24; 2160p resolution). The display was placed 3 m away from the participant, centered on their midline, perpendicular to the sagittal plane.
Study set-up and protocol. A) Participant in the arm support chair holding the squeeze ball in their right hand. The vibrotactile display is placed on their left arm. The participant wears custom goggles that block the direct view of the moving arm and hand, while allowing a full view of the display screen. B) Visual workspace showing a 5 × 5 × 5 grid of targets (presented on a black background for the experiment). Targets are placed at 0%, 47.5%, and 95% of each calibrated hemispace. The z-dimension is represented by circle size, such that the target diameter decreased linearly as the target's distance from the body increased (i.e., smaller targets are further away from the torso). The centermost ring represents the “home” position . A target is shown as a solid green circle. During trials with visual feedback, a cursor (solid black circle here; white circle for the experiment) is shown corresponding to the position of the hand in the reachable workspace. Additionally, out-of-bounds indicators were displayed when the participant exceeded 105% of the calibrated hemispace (red bars for ± x/±y and a red cursor for ± z). C) Vibration motor placement on the stationary (ipsilesional) arm. Vibration motors are placed on either the left or right arm, as shown. D) Protocol of this 21-session study, with feedback conditions labeled for each block. We required a minimum interval of 24 hours between sessions. During training sessions (V), goal-directed reach-to-grasp actions are performed using vibration feedback without concurrent visual feedback of either the cursor or the contralesional limb; visual cursor motion is provided after the squeeze, thereby enabling the participant to correct target capture errors. Testing sessions (visual and T) do not include this after-squeeze visual feedback (and consequent corrections).
Visual Display
The visual display showed a 5 × 5 grid of targets in the ± x / ±y plane of the screen, with 5 concentric rings representing the additional dimension of reach depth (±z) at each location, yielding 125 potential visual targets within the reachable workspace (Figure 1B). The center-to-center inter-target distance along the x and y planes was 11 cm, and the ring size (z dimension) ranged in diameter from 2 cm to 9 cm. The central “home” position, representing the origin of the workspace, was shown in the middle row, column, and depth of the visual display. This home position divided the workspace into six hemispaces, each comprising the half of the workspace on each side of a given axis (i.e., +x, -x, +y, -y, +z, and -z). The visual display was normalized to each participant's unique reachable workspace such that targets were placed at 0%, 47.5%, and 95% of each calibrated hemispace (calibration process described later in Experimental Protocol). Of the 125 visual targets, 100 were used during reach-to-grasp training with supplemental kinesthetic vibrotactile feedback (Figure 1D, Block V), whereas 25 targets were set aside for testing purposes only (Figure 1D, Block Visual and Block T).
During some reach-to-grasp movements, the real-time position of the hand (inferred from the Optotrak markers attached to the squeeze bulb) was used to calculate the position of a visual cursor, which was projected onto the workspace spanned by the grid of visual targets ( Figure 1D, Block Visual and Block V inter-trial corrections). During these movements with visual cursor feedback, out-of-bounds indicators were presented when the participant reached beyond 105% of the calibrated hemispace (Figure 1B). We included this occasional visual cursor feedback during training sessions (Figure 1D, Block V inter-trial corrections) to allow participants to realign their inference of hand position from the real-time vibrotactile feedback to the ground truth feedback of the hand's position in peripersonal space provided by vision.
Vibrotactile Display
We attached a “vibrotactile display” to the stationary ipsilesional arm to provide supplemental vibrotactile feedback regarding the real-time position of the moving, contralesional hand (Figure 1A). The vibrotactile display comprised six eccentric rotating mass vibration motors (Precision Microdrives Inc., London, UK; model 310–117), which have an operational frequency range of 60 to 230 Hz and a covarying amplitude range of 0.2 to 2.2 G. Each vibration motor was affixed directly to the skin of the stationary ipsilesional arm with elastic fabric bands or Transpore tape (3 M, Transpore 1527-1; Figure 1C). With reference to anatomic position, the placement of the six vibration motors included: (i) C7 dermatome on the dorsal hand, about 5 cm proximal to the middle finger knuckle; (ii) C8 dermatome on the ventromedial forearm, approximately 5 cm proximal to the ulnar styloid process; (iii) C6 dermatome on the ventrolateral forearm, approximately 5 cm distal to the antecubital fossa; (iv) T1 dermatome on the posterior arm, about 5 cm proximal to the olecranon; (v) C4 dermatome superior to the clavicle, on the subclavian triangle; (vi) and C7 dermatome on the scapula, about 3 cm inferior to the scapular spine [cf. (Whitman & Adigun, 2022)]. The spacing between motors was greater than 6 cm at all sites, thereby avoiding undesirable mechanical crosstalk between adjacent sites (Nolan, 1982; Shah et al., 2019).
The vibrotactile display provided limb state encoding such that all motors behaved the same way for deviations from the central “home” position along their individually-assigned hemispace (±x, ±y, ±z). Specifically, no vibration was present at the home target (i.e., the origin of the workspace). The motors maintained zero vibration within a 1 cm distance of that central point in all directions. Outside of this small dead-zone representing the home target, the intensity of vibration increased linearly before saturating with maximal vibration at 105% of the hemispace. This activation pattern, which was common to all six vibration motors within their respective hemispace, is described in Figure 2A.
An example of vibrotactile motor activations. A) Activation of one motor with respect to deviations from the center of the home position within its respective hemispace. This activation pattern applies to all six motors (in their respective hemispaces). Note that the motor activations are zero for deviations less than 1 cm from the home position, whereas the activation function linearly increases for deviations greater than 1 cm, with the slope depending on the size of the participant's reachable workspace (in units of % workspace). B) Example of a hand path from one selected trial. The participant's starting position is denoted by “1” in the upper left quadrant of the displayed workspace, and the ending position is represented by “4” on the right side of the middle row of targets. “Snapshot” tracings of the participant's hand path across time are shown by black rings with two selected points highlighted. It should be noted that a cursor is not shown during vibrotactile reaching trials; these snapshots are depicted here for explanatory purposes only. The z-dimension is represented by circle size, such that the target diameter decreased linearly as target distance from the body increased (i.e., smaller targets are further away from the torso). C) Activation of the six different motors across time for the trial presented in panel B; vertical lines with matching labels correspond to the four different points highlighted in panel B. From top to bottom, the pairs are shown for x, y, and z deviations. Solid lines: deviations into the positive hemispace. Dashed lines: deviations into the negative hemispace.
Consider Figure 2B-C for an example of how the vibrotactile feedback was provided. In Figure 2B, the participant's starting position is denoted by the label “1” and the ending position by the label “4”. Tracing of the participant's hand path across time are shown by black rings, with two additional points highlighted. Figure 2C shows the activation of the six different motors across time. At the start position, the -x, +y and -z motor would be active. As the participant reaches through point “2”, they would cross the xz-plane (moving from the upper hemispace to the lower hemispace), therefore switching activation from the + y motor to the -y motor. While reaching between points “2” and “3”, the hand crosses the yz-plane (moving from left to right), therefore switching activation from the -x motor to the + x motor. Around point “3”, the participant crosses the xy-plane (moving closer to the torso, denoted by the increase in black ring diameter from smaller to larger than the middle ring) therefore switching activation from the -z motor to the + z motor. At the end of motion, neither of the y motors would be active since the hand is centered on the xz-plane. It should be noted that a visual cursor would not be displayed during vibrotactile feedback trials.
Experimental Protocol
In our earlier work with a cohort of younger, neurologically-intact participants, we observed a plateau in training-related improvements of spatial accuracy and temporal efficiency after approximately five hours of training (30 minutes of training per day over the first 10 out of 20 days of training; Shah et al., 2023). Here, we allowed our participants to engage in about the same total amount of training to accommodate the possibility that survivors of stroke might exhibit somewhat slower rates of performance improvements. Our design for this longitudinal study also sought a balance between adequate training time vs. having an excessive number of training sessions, which could conceivably have a negative impact on recruitment and retention. Thus, as depicted in Figure 1D, each participant completed 21 experimental sessions, with each session performed on a separate day (i.e., with a minimum break of 24 hours between sessions). The average inter-session interval was 5.3 ± 4.2 days (mean ± 1 standard deviation).
The first session involved clinical testing (described in Clinical Evaluations and Participant Inclusion). The second session started with a calibration procedure wherein the span of the participant's reachable workspace was measured in all three cardinal directions; this information was used to establish the participant's individual, unique workspace. For this calibration process, participants were asked to reach to eight locations that formed a rectangular prism within their peripersonal space. The width of this prism did not exceed shoulder width; the height of the prism ranged from just above the participant's seated legs to just below shoulder height; the depth of the prism ranged from the participant's chest to the furthest outward, cross-body reach while the participant's torso maintained contact with the chair. The geometric center of the prism defined the origin of this workspace and the location of the central “home” target. On average, the workspace spanned x: 23.2 ± 7.5 cm, y: 19.3 ± 6.0 cm, and z: 16.1 ± 5.5 cm. Additionally, during calibration, an individualized squeeze threshold was defined as 50% of the participant's maximal squeeze strength.
Sessions 2 to 21 required participants to make reach-to-grasp movements using their contralesional arm and hand. Each of these sessions started with a familiarization/reacquaintance phase wherein participants were allowed to briefly explore the correspondence between the visual and vibrotactile displays as they moved the cursor with their hand. The vibrotactile display was verbally described to participants as they were instructed to reach to specific targets aligning with selected vibration motor activations. During this phase, participants were frequently asked to report which motor(s) were activated in the moment. Adjustments were made to individual vibration motor locations (i.e., by moving each by 2 or 3 cm within the same dermatome) as needed until the participant could reliably detect and report vibrations from all six motors. Participants were then encouraged to explore the vibrotactile display by making self-guided reaching movements until comfortable with the mapping between the visual workspace, their reachable workspace, and the sensorimotor space defined by the vibrotactile display. This familiarization/reacquaintance procedure typically took between 2 and 5 minutes to complete.
Participant Instructions
The main part of the experiment required subjects to make reach-to-grasp movements between targets presented one at a time on the visual display screen (one reach-to-grasp target capture per trial). At the start of each trial, the target (a green filled circle; Figure 1B) appeared on the screen for two seconds as a preparatory cue; after this period, an auditory “GO” signal indicated that the participant should reach to the target with their contralesional hand. Participants were instructed to “capture the target as quickly and accurately as possible” and then to squeeze the ball in their contralesional hand upon arriving at the perceived target location. This squeeze event was recorded as the target capture event regardless of whether or not the hand actually captured the desired target. Participants had 60 seconds to initiate the squeeze event before a given trial automatically ended and the next trial began.
Testing Sessions
The study protocol included two testing sessions designed to assess each participant's baseline and final performance using supplemental vibrotactile kinesthetic feedback without concurrent visual feedback to guide real-time control of the contralesional arm post-stroke (Figure 1D, Block T). Testing sessions involved two blocks of 25 reach-to-grasp movements selected from the set of generalization targets excluded from the set of training targets. One of these blocks, referred to as the intrinsic feedback block, presented no additional feedback of hand position such that participants had to rely solely on intrinsic proprioception to reach to the goal target. The other block (vibrotactile feedback) provided participants with additional vibrotactile feedback of hand position relative to the center of the workspace. The first testing session additionally included a block of 25 trials wherein a real-time visual cursor representing the participant's hand position was represented on-screen (Figure 1B, visual cursor; Figure 1D, Block Vision). The baseline test was conducted during Session 2 to assess the participant's initial reach-to-grasp performance with each type of feedback (i.e., visual, intrinsic, and vibrotactile). The final test was conducted during Session 21, which occurred after 18 sessions (9 hours) of vibrotactile feedback training. Final testing only included the intrinsic and vibrotactile feedback conditions.
Training Sessions
Training sessions required participants to make reach-to-grasp actions with concurrent supplemental vibrotactile feedback between targets drawn from a set of 100 targets that excluded the 25 testing targets. Training sessions were divided into three 10-minute sections, with 2-minute breaks between each section. Participants were instructed to perform the reach-to-grasp task at a comfortable, self-selected pace. After indicating perceived target capture by squeezing the squeeze ball, participants were provided real-time visual cursor position feedback along with the vibrotactile feedback and they were instructed to correct hand positioning errors relative to the intended target. During these after-target corrections, participants were encouraged to note how their perception of hand location changed as its representations moved simultaneously within the vibrotactile and visual displays. Training sessions ended after 30 minutes such that subjects did not have to complete all 100 possible target capture actions within each session.
Subjective User Experience Surveys
After the last testing block of the last experimental session, participants were asked to complete a set of standard surveys quantifying subjective user experience in three domains (i.e., system usability, motivation, and user satisfaction), as well as to respond to a set of open-ended questions (see Supplemental Material online). The System Usability Scale (SUS) served as the primary measure of system usability (Brooke, 1996); the SUS assessed the usability of the vibrotactile feedback within the context of enhancing the accuracy and efficiency of goal-directed reaching movements. Participants responded to ten prompts using a 5-option Likert scale ranging from “Strongly Disagree” to “Strongly Agree”. Total SUS survey scores greater than or equal to 68 on a 100-point scale indicate passable usability. The Intrinsic Motivation Inventory (IMI) assessed the extent to which participants perceived the supplemental vibrotactile feedback to be motivating via six dimensions of the original survey: interest/enjoyment, effort/importance, value/usefulness, perceived competence, perceived choice, and perceived pressure/tension (McAuley et al., 1989). Participants responded to thirty-seven prompts using a 7-option Likert scale ranging from “Not-At-All True” to “Very True”. IMI scores greater than or equal to a value of 4 on the 7-point scale indicate favorable motivation to use the system; the only exception to this rule is the pressure/tension subsection, where lower scores are desired to indicate that the participant did not feel pressure or tense when using the system. Both the SUS and IMI use variations of the same questions (including reverse wording) to increase response reliability and to reduce response bias. Finally, the Quebec User Evaluation of Satisfaction with assistive Technology 2.0 (QUEST) assessed user satisfaction with six aspects of the supplemental vibrotactile feedback system: weight, safety, ease of use, comfort, effectiveness, and instructions for use (Demers et al., 2000, 2002). QUEST is comprised of two parts. For the first part, participants responded to six prompts using a 5-option Likert scale ranging from “Not-Satisfied-At-All” to “Very Satisfied”, addressing their satisfaction with the six items. Scores equal to or greater than a value of 3 on a 5-point scale indicated positive user satisfaction with the assistive device. The second part of QUEST asked participants to identify the three most important aspects (from the six listed above) contributing to their perceived satisfaction with the feedback system.
Subjects were also provided with an opportunity to provide an open-ended written response to the questions:
What did you like/dislike about the supplemental vibrotactile feedback device? How has your ability to use the supplemental vibrotactile feedback device to help reach the targets changed over the course of the study? How well do you think the supplemental vibrotactile feedback device helped you reach the targets? What strategies did you use to interpret/utilize the supplemental vibrotactile feedback? What did you like/dislike about this training? How has your participation in this study affected how you use your arm in daily life?
Data Processing
Squeeze-bulb pressure data were filtered in real-time using a 5-sample sliding-window average prior to being compared to the participant's individual squeeze threshold. A squeeze event was identified and documented whenever the average squeeze pressure transitioned from below threshold to above threshold. Trials where 60 seconds elapsed without a squeeze event were removed from analysis (8% of collected trials).
Motion capture data from each reach-to-grasp movement were examined off-line for missing marker events that could occur due to occlusion by the participant's fingers or thumb. Missing marker data were interpolated using a cubic spline. The resulting kinematic time series were used to compute three primary performance measures: target capture error (a measure of target capture accuracy), average movement speed (a measure of temporal efficiency), and hand path length ratio (a measure of spatial efficiency). Target capture error was defined as the distance between the center of the target and the participant's hand position in the calibrated user's reference frame at the onset of the participant's squeeze event (i.e., the moment of perceived target capture). To facilitate comparisons of performance across participants with different reachable workspace volumes, target capture error is reported as a percentage of the normalized workspace, which was computed by dividing the total Euclidean distance between the user's end position and target by the Euclidean span of the participant's workspace. Average movement speed was defined as the total distance traversed by the hand from its location at the start of the reach (i.e., the moment the hand first exceeded 10% of its peak velocity on a given trial) to its location at the onset of the participant's squeeze event, divided by the time elapsed from movement onset to the squeeze event onset. Average movement speed was selected over target capture time as a measure of temporal efficiency to account for differences in inter-target distance across trials. Finally, hand path length ratio was defined as the actual distance traversed by the hand from movement onset to squeeze onset divided by the straight-line distance between the hand's starting position and ending position; an optimally executed straight-line reach would have a hand path length ratio of 1.0.
Statistical Testing of Hypotheses
We evaluated the kinematic performances of individual survivors of stroke as a case series to test two main hypotheses. The first poses that extended, multi-session training with 3-dimensional vibrotactile feedback and without concurrent vision would lead to improved reaching accuracy and efficiency for survivors of stroke. We used a case series approach rather than a cohort analysis to test this first hypothesis due to the wide range of sensory and motor impairments presenting in our small cohort. Thus, for each participant, we used a within-subjects analysis comprising four planned t-tests applied to each of the three trial-based kinematic performance measures (target capture error, average movement speed, and hand path length ratio): 1) between the first and last
The second hypothesis poses that participants would find the vibrotactile feedback system useful, motivating, and satisfying to use. Because we did not expect user experience to depend on the degree of sensorimotor impairment, we used 95% confidence intervals from the cohort to compare each survey item to its corresponding “positive experience” threshold (described in 2.10. Subjective user experience surveys). The questionnaire soliciting open-ended feedback was not used to generate a score intended for statistical hypothesis testing; rather, it was used to gather subjective feedback about the study protocol and potential effects outside the lab sessions. All statistical testing was performed in R Studio 2024.12.0.
Results
All participants were alert throughout each experimental session. Participant S01 did not undergo baseline testing due to a hardware failure that impacted data collection that day; this participant was excluded from statistical hypothesis tests that relied on baseline data (i.e., tests 2 and 3 described in section Statistical Testing of Hypotheses).
Effects of Vibrotactile Feedback Training on Primary Measures of Reach Accuracy and Efficiency
Figure 3 presents hand paths from four blocks of trials performed by two selected subjects. Each panel shows hand path data from a single test block (25 trials) projected onto the three two-dimensional planes spanning the reachable workspace; each plane aligns with one combination of two cardinal axes from the 3-dimensional vibrotactile display. The top panels show hand path data from participant S07, who exhibited the greatest improvement in mean target capture accuracy between baseline and final testing blocks performed with supplemental vibrotactile feedback (see also Figure 4A, TESTING, dash-dot line). As can be noted in Figure 3, this participant's movements were biased above the workspace midline (i.e., a positive y-axis bias) and away from the body (a negative z-axis bias) during baseline testing under both feedback conditions. After 9 hours of training with supplemental vibrotactile feedback, these biases were largely eliminated under both conditions. This participant also became more efficient in their target captures in the sense that hand paths appeared to be shorter after training compared to before training under both feedback conditions; this can be inferred from the decreased density of hand paths in the final testing blocks (see also Figure 4C, TESTING; dash-dot lines).
Hand position plots from all 25 trials per block in both testing sessions with intrinsic and vibrotactile feedback from the subject with the most improvement in vibrotactile testing accuracy (S07) and the subject with the least improvement in vibrotactile testing accuracy (S05). In each panel, the 3-dimensional hand movements are decomposed and projected into 2-dimensional components. The results were normalized to the participant's unique reachable workspace. Targets were placed at 0%, 47.5%, and 95% of each calibrated hemispace.
Kinematic results for each participant: A) Euclidean target capture error, B) Euclidean average movement speed, and C) total hand path length ratio. Target capture error is the percent workspace difference between the center of the target and the final hand position. Average movement speed is the distance traversed divided by time elapsed from the “GO” cue to the squeeze event. Hand path length ratio is the actual distance traversed divided by the optimal distance from the hand's start to end position. Individual scores with visual feedback are represented by dots, and learning trends for other feedbacks are illustrated by lines. Cohort mean values are shown in the margins adjacent to the individual-participant data plots. The training sessions included after-squeeze visual and vibration feedback for inter-trial corrections. Testing sessions did not include after-squeeze corrections. Dash-dot lines: S07. Dotted lined: S05. Open circle: cohort mean with visual feedback. Open square: cohort mean with vibrotactile feedback training. Open triangle: cohort mean with vibrotactile feedback testing. Open diamond: cohort mean with intrinsic feedback. Error bars: ± 1 standard deviation of the cohort mean.
By contrast, the bottom panels of Figure 3 show hand path projections from participant S05, who exhibited the least improvement in mean target capture accuracy from baseline to final testing with vibrotactile feedback (see also Figure 4A, TESTING, dotted line). This participant's baseline movements were also biased above the workspace midline (i.e., a positive y-axis bias) but toward the body (a positive z-axis bias) under intrinsic feedback. These biases persisted under intrinsic feedback even after 9 hours of training with supplemental vibrotactile feedback. Although the biases were not seen in baseline movements with supplemental vibrotactile feedback, they did appear after training, suggesting the possibility that this participant may have come to ignore the vibrotactile feedback over time.
Figure 4 presents average performances from each participant in selected trial blocks for each of this study's three primary outcome variables: target capture error (Figure 4A), average movement speed (Figure 4B), and hand path length ratio (Figure 4C). Target capture error was our primary measure of reach-to-grasp accuracy. When provided with real-time visual cursor feedback during reaching, all participants demonstrated an ability to perform the reach-to-grasp task with relatively low target capture error values (range: 3.79% to 13.84%). Target capture error increased markedly when visual feedback was initially removed (compare performance in the two baseline testing blocks to performance in the visual block). During training, all but one of the seven participants tended to decrease target capture errors to some extent from the first to the last training block performed with real-time vibrotactile kinesthetic feedback and post-reach visually-guided corrections. Training had mixed impact on ability to use the vibrotactile feedback to improve movement accuracy in the final testing blocks (i.e., in the absence of post-reach visually-guided corrections): whereas 4 of 6 participants appeared to decrease
Average movement speed was our primary measure of temporal efficiency of reach-to-grasp actions. When provided with visual cursor feedback during reaching, participants varied considerably in average movement speed (range: 2.8 cm/s to 13.4 cm/s; Figure 4B). Participants were also heterogeneous in the impact of training on average speed. Whereas some participants appeared to increase average speed from the first
To determine whether the heterogeneity observed in Figure 4 could reflect the possibility that some participants responded favorably to training with the supplemental vibrotactile kinesthetic feedback whereas others did not, we performed within-subject analyses for each of the three trial block contrasts depicted graphically in Figure 4 (i.e., vibrotactile training, vibrotactile testing, and intrinsic testing), and for a contrast comparing performance in the two final testing blocks (i.e., supplemental vibrotactile feedback vs. intrinsic only feedback). Table 2 presents analyses of individual participant performance changes between the first and last
Individual Performance Changes [Within Subject Mean (SD)] Between First and Last Vibrotactile Training Sessions (Hypothesis 1, Test 1).
Table 3 presents analyses of performance changes between baseline and final
Individual Performance Changes [Within Subject Mean (SD)] Between Baseline and Final Vibrotactile Testing Sessions (Hypothesis 1, Test 2).
Table 4 presents analyses of performance changes between baseline and final
Individual Performance Changes [Within Subject Mean (SD)] Between Baseline and Final Intrinsic Proprioception Testing Sessions (Hypothesis 1, Test 3).
We performed a final test of hypothesis 1 wherein we analyzed within-subject performance differences between movements made during final
Individual Differences [Within Subject Mean (SD)] Between Intrinsic and Vibrotactile Feedback for the Final Testing Sessions (Hypothesis 1, Test 4).
Subjective User Experience of Supplemental Vibrotactile Feedback and Training
Subjective user experience was assessed using standard surveys and open-ended questions after each participant's 21st experimental session (Figure 5). In contrast to the marked difference in individual performance in the use of supplemental vibrotactile feedback during reaching (described above), survey responses were largely consistent within our small cohort of participants. The cohort's mean SUS score resided above the threshold of positive system usability (a value of 68) although the 95% confidence interval dipped slightly below this threshold (mean: 76. 07; 95% CI: 63.35–88.79). All dimensions of the IMI had their means and 95% confidence intervals fully above the corresponding threshold value (threshold: 4; lowest CI: 4.53). The QUEST also yielded scores with each satisfaction items’ 95% confidence interval residing above the threshold for positive satisfaction (threshold: 3; lowest CI: 3.59). With regards to which items were reported as most impactful for satisfaction, the top three were device instructions, comfort, and effectiveness. These data support the hypothesis that participants would find the vibrotactile feedback useful, motivating, and satisfying to use.
Cohort subjective user experience results for the A) System usability scale score, B) Intrinsic motivation inventory score, C-D) Quebec user evaluation of satisfaction with assistive technology scores (C: mean score, D: frequency and score for each satisfaction item). Gray circles: individual scores. Black dashed lines: threshold for positive user experience. Error bars: 95% confidence interval of the mean. *Tension scores were subtracted from 8 to visually have passable scores above the threshold.
Discussion
This case series report described a set of experiments that evaluated the ability of seven stroke survivors to benefit from multi-day 3-dimensional kinesthetic vibrotactile training as a means to enhance the accuracy and efficiency of goal-directed reach-to-grasp actions performed in the absence of visual feedback. A motion capture system converted real-time contralesional hand position in a Cartesian frame of reference into spatiotemporal patterns of vibrotactile feedback provided to the non-moving, ipsilesional arm. Each participant demonstrated competence by performing the reach-to-grasp task reasonably well when provided concurrent feedback of a visual cursor representing hand location within peripersonal space. As expected, target capture errors increased markedly (i.e., accuracy decreased) for all participants when visual feedback was initially removed, whether or not they were provided the supplemental feedback. Participants then underwent 9 hours of reach-to-grasp training under conditions that encouraged them to learn a mapping from hand position to patterns of vibrotactile feedback. Within-subject analyses comparing post-training performances both with and without the supplemental vibrotactile feedback to the corresponding baseline conditions found heterogeneous patterns of training-related improvements in reach-to-grasp accuracy and efficiency under both testing conditions for some participants. Within-subject comparisons of final
Design Considerations for Supplemental Feedback Systems
The development and evaluation of practical supplemental sensory feedback systems is important because they have potential to mitigate functional impairment via mechanisms of sensory augmentation or replacement. Previously, supplemental feedback has been used to translate visual information into electrotactile stimuli delivered to the tongue for those with visual impairments (Bach-Y-Rita et al., 1969; Collins & Bach-y-Rita, 1973; Nau et al., 2015), to promote motor learning while playing an instrument or sports (Lieberman & Breazeal, 2007; Ruffaldi et al., 2009; van der Linden et al., 2011b), to augment motor performance in complex tasks such as robot-assisted surgery (Abiri et al., 2019; Kent & Rossa, 2022; Schoonmaker & Cao, 2006), and to improve sensorimotor control by exciting corticospinal pathways contributing to the regulation of movement and/or reflex activity (Conrad et al., 2011).
A growing body of research has sought to directly encode important aspects of movement performance into supplemental vibrotactile stimuli that either are intended to be used indefinitely to augment impaired sensory function (Shull & Damian, 2015) or to teach desirable skills [e.g., through a guided re-weighting of intact afferent signals; (Sienko et al., 2018)] that are intended to persist after the vibrotactile stimuli are removed (Bark et al., 2011; Holden & Todorov, 2002; Lieberman & Breazeal, 2007; van der Linden et al., 2011a). We hypothesized that our small cohort of stroke survivors would use the supplemental feedback system like a prosthesis, such that extended training with 3-dimensional vibrotactile feedback would lead to improved in-the-moment accuracy and efficiency of reach-to-grasp actions due to the real-time kinesthetic information encoded within the feedback. This was not supported by the analyses reported here (Table 5). Instead, we observed that many participants who showed significant improvement with vibrotactile feedback (Table 3) also demonstrated improvements without the supplemental feedback (Table 4), suggesting that use of the device in this study acted to train more effective use of residual afferent pathways. Further anecdotal support for this retraining effect derives from individual participant reports of increased contralesional function even after the vibrotactile display was removed. Similar findings of persistent improvement in accuracy for post-training reaches performed without vibrotactile feedback was also observed after short-term training with neurologically-intact younger adults (Risi et al., 2019). Taken together, these findings suggests that the present system could be usefully integrated within a therapeutic program of mass practice to promote improved movement accuracy and increased hemiparetic arm use. Indeed, the current state-of-the-art approach to increasing hemiparetic arm use is an intervention called constraint induced movement therapy (CIMT), which uses high-repetition task-training of the more involved upper limb while also constraining the less involved side. CIMT requires 6 hours of training per day for 2 weeks under the guidance of a trained clinician (Taub et al., 2013). CIMT can yield clinically meaningful improvements in more involved upper limb function that can persist long after intervention (Wolf et al., 2008). Unfortunately, widespread clinical adoption of CIMT has been limited by the intensive time demands it places on clinicians (Pedlow et al., 2014). As such, there is unmet need for interventions that engage the more involved upper limb in recovery while also minimizing clinician oversight. The vibrotactile feedback system may provide an accommodation to this need. To the extent that the post-training effects presented in Table 4 reflect a carry-over effect, along with our finding that three of the participants also improved reach accuracy during post-training with vibrotactile feedback (Table 3), indefinite use of the vibrotactile feedback system could be considered, at least for some survivors of stroke.
Training-Related Changes on Primary Measures of Reach Accuracy and Efficiency
Recent related studies conducted with neurologically-intact young adults have found that short-term vibrotactile limb state feedback training can yield persistent and beneficial effects wherein participants learn to construct a mapping between visual targets in peripersonal space and the novel geometric representation of that space within the vibrotactile display, and then use it to effectively improve the accuracy of goal-directed reaching movements performed without visual feedback (Mazorow et al., 2024; Rayes et al., 2023; Risi et al., 2019). Those studies found that healthy young adults were able to significantly reduce target capture errors after just one training session such that movements performed with vibrotactile feedback yielded greater accuracy than those performed with intrinsic proprioceptive feedback alone, even after accounting for the training-related improvements in proprioceptively-guided reaching. Improvements in accuracy came at the cost of decreased temporal and spatial efficiency of movement, suggesting that the real-time integration of vibrotactile kinesthetic feedback into the ongoing control of goal-directed movements is a cognitively demanding task that requires overt attention (Shah et al., 2018).
Previous studies that included stroke survivors found that they also were able to use supplemental vibrotactile feedback to enhance the accuracy of movements made in 1-dimension (Tzorakoleftherakis et al., 2015). However, only some survivors were successful with using supplemental feedback to enhance the accuracy of 2-dimensional movements (Ballardini et al., 2021). In the current study, we observed that only three of our six participants with baseline data improved the accuracy of 3-dimensional movements after vibrotactile feedback training during
Two factors might explain why vibrotactile feedback was not as effective in our small cohort of stroke survivors. First, neurological impairment following stroke may have compromised neural circuitry supporting the form of spatial learning required to construct and use an accurate mapping between visual targets in peripersonal space and patterns of vibratory feedback within the vibrotactile display. If so, such deficits should be evident in standard clinical tests of spatial reasoning, such as the Trails A and Trails B tests from clinical psychology (Bowie & Harvey, 2006; Reitan, 1955). Future studies proposing to use supplemental kinesthetic feedback to improve reach-to-grasp accuracy should quantify the capacity for spatial reasoning as a potential cofound to the effective use of supplemental vibrotactile kinesthetic feedback. Second, it is well known that aging compromises the ability to process information and translate it into physical action [i.e., psychomotor processing speed; (Smith, 1973)]. It is therefore possible that the complexity of 3-dimensional movements and the use of 6 vibration motors to provide the supplemental kinesthetic feedback could have overwhelmed the majority of our participants; indeed, the three participants demonstrating potential improvements in reach accuracy with vibrotactile feedback were the youngest three participants in our study (mid 50s or younger). The heterogeneity we observed in the ability of our participants to capitalize on supplemental vibrotactile feedback to enhance reach accuracy supports both the need to identify which survivors of stroke might benefit from the intervention, and the need to individualize the device and training plan to each participant.
We also note that one of the three participants who improved reach accuracy in our study did so at the expense of spatial efficiency. This trade-off is consistent with a similar trade-off observed in previous studies (Mazorow et al., 2024; Rayes et al., 2023; Risi et al., 2019). This could indicate that more time may be needed to internalize 3-dimensional supplemental feedback before efficiency starts to improve. However, a related study wherein healthy young adults trained to make horizontal planar reaching movements with supplemental vibrotactile kinesthetic feedback found that performance reliably improved, and then plateaued, after only about 5 hours of training distributed over approximately 10 days (Shah et al., 2023). Thus, we do not think it likely that additional hours of training would confer great benefit to participants who did not demonstrate at least some degree of improvement over the 9 hours of training employed here.
Effects of Participant Impairment Level on Primary Measures of Reach Accuracy and Efficiency
We noted no clear relationship between training-related changes in reach-to-grasp performance and any of the clinical assessment scores summarized in Table 1, although it is important to keep the small size of our sample in mind. Three participants had moderate to severe motor or sensorimotor deficits as determined by clinical testing. S04 (sensory and motor deficits) showed significant improvements in accuracy and spatial efficiency of
Mild cognitive impairments in three participants (S02, S04, and S06) did not appear to preclude performance improvements over the course of the study. During
Subjective User Experience During Supplemental Vibrotactile Feedback Training
Subjective user experience is important to consider when developing technology-based interventions because even the most effective intervention will have little value if patients perceive the intervention to be difficult to use, unmotivating, or otherwise unsatisfactory. In one earlier study, Rayes et al. (2023) compared the utility and usability of two different vibrotactile feedback encoding schemes during goal-directed reaching in the horizontal plane with healthy young adults. One feedback scheme communicated changes in shoulder and elbow joint angles, whereas the other communicated changes in hand position in a Cartesian coordinate frame, as in the current study. Using the same surveys as the current study, the authors reported passable scores for motivation, usability, and satisfaction for limb state feedback, as was also reported here (Figure 5). Similarly positive user experience scores were reported in a study of 3-dimensional vibrotactile feedback training with healthy young adults (Mazorow et al., 2024). In both studies, the younger participants selected “ease of use”, “comfort”, and “effectiveness” as the top three criteria for device satisfaction on the QUEST survey. Although two of these criteria overlapped with those selected by survivors of stroke in the current study (i.e., comfort and effectiveness), the survivors ranked “instructions” much higher than healthy young adults, suggesting that audience age and neurological status should be considered when implementing future designs of vibrotactile training protocols.
Open-ended feedback provided by many participants indicated that they enjoyed the challenge of the task, and that participating in the study had increased contralesional function and awareness outside of the study sessions. Additionally, a few stated that the task felt like a puzzle or game and kept them cognitively challenged enough to keep interest and be more attentive than traditional physical or occupational therapy sessions. Taken together, our reports of subjective user experience show promise in that these users would likely commit to using 3-dimensional vibrotactile feedback in an assistive device if it were proven to be effective, comfortable and easy to use, with clear, easy-to-understand instructions provided.
Some inconsistencies were noted between written responses to the open-ended questions and the kinematic performance of certain participants. S05 reported growing confidence in completing the task with the vibrotactile feedback despite exhibiting worsening accuracy during
Limitations
This study has several limitations. This case series includes only a small number of participants who had a broad range of sensory, motor, and cognitive impairments. A larger cohort with similarly-impaired sensation and residual motor capacity is needed to definitively assess the benefits of extended training on the use of supplemental kinesthetic feedback training with a vibrotactile display. To do so, however, would add considerable difficulty to participant recruitment and retention. As a first step, this study chose to recruit participants with a broader range of sensorimotor impairment to facilitate recruitment. To promote retention, we additionally allowed flexibility in the scheduling of sessions, which yielded variability in the amount of time between sessions (5.3 ± 4.2 days). We cannot discount that this flexibility could have led to a decrease in the retention of learning from session to session. While a stricter scheduling regime might lead to greater performance improvements across sessions, it could also negatively impact participant retention within the longitudinal study.
Another limitation is our inability to track whether participants continued to utilize the supplemental vibrotactile feedback throughout each session. While everyone demonstrated vibration discrimination during the familiarization/reacquaintance phase, as well as at the start of each 10-minute section of training (i.e., after breaks), it is possible given the complexity of the supplemental vibrotactile feedback that participants’ attention to the feedback could have waned as each section of training progressed. This would ultimately affect the performance improvements observed during experimental blocks performed with the vibrotactile feedback. The open-ended questionnaire was used in an attempt to address this limitation, specifically asking “What strategies did you use to interpret/utilize the supplemental vibrotactile feedback?”. All participants described strategies to engage with the device, although S06 did mention that she sometimes noticed her attention waiver from the vibrations. This may account for why this participant showed similar improvement in average movement speed across
We also acknowledge that using hand path length ratio as a measure of spatial efficiency oversimplifies reach-to-grasp movements by neglecting to consider subtle differences in control based on the directionality of movement, e.g., by treating the initial ballistic phase of movements equivalent to the end-of-movement fine-corrections around the goal target, when in fact these two aspects of control may be distinct (Ghez et al., 2007; Scheidt et al., 2011; Scheidt & Ghez, 2007).
Finally, this study did not include a control group performing the same amount of reach training without vibrotactile feedback. Absent such a control, it is impossible to attribute performance gains during training to the vibrotactile device rather than to general practice effects. However, our experimental design does permit within-subject comparisons between post-training
Conclusion
Even though participants generally perceived the supplemental vibrotactile feedback system to be useful, intrinsically motivating and satisfying to use, the results of this case series suggest that only a subset of stroke survivors may actually accrue benefits in terms of improved reach-to-grasp accuracy while using supplemental vibrotactile kinesthetic feedback. Moreover, within-subject analyses suggest that the increased accuracy and efficiency of reaching movements may result more from the performance of 9 hours of focused reach training with the supplemental feedback, rather than from the moment-by-moment integration of the information conveyed by the feedback system into the ongoing control of movement. Future studies should seek to identify those participants who might benefit most from this form of supplemental kinesthetic feedback (e.g., by identifying sensorimotor or neuropsychological clinical tests and/or neuroimaging biomarkers that can serve as predictors of baseline performance and improvements in reach-to-grasp performance with training).
Supplemental Material
sj-docx-1-rnn-10.1177_09226028261429797 - Supplemental material for Utility and User Experience with 3-Dimensional Vibrotactile Kinesthetic Feedback for Improving Reach-to-Grasp Accuracy and Efficiency After Stroke: A Case Series
Supplemental material, sj-docx-1-rnn-10.1177_09226028261429797 for Utility and User Experience with 3-Dimensional Vibrotactile Kinesthetic Feedback for Improving Reach-to-Grasp Accuracy and Efficiency After Stroke: A Case Series by Rachel N Mazorow, Ramsey K Rayes, Joana Flores, Gina H Guidarelli, Kimberly D Bassindale and Robert A Scheidt in Restorative Neurology and Neuroscience
Footnotes
Acknowledgements
The authors thank Dr. David Reinkensmeyer for use of the T-WREX arm support device, Dr. Kristine Mosier for use of an Optotrack 3020 motion measurement system, St. Camillus Life Plan Community and Jenny Zimpel for providing an off-campus location for our sessions, and Marquette University's Aging, Imaging, and Memory (AIM) Lab for conducting neuropsychological testing.
ORCID iDs
Ethical Considerations
This study received ethical approval from the Institutional Review Board of Marquette University (approval: HR-3303) on March 19, 2020.
Consent to Participate
All participants provided written informed consent prior to participating.
Consent for Publication
All participants provided written informed consent to publish the data collected for this study.
Author Contributions
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the NIH Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) grant R15 HD093086-02 and by the Rev. John P. Raynor S.J. Research Chair at Marquette University. R.N. Mazorow received support from Marquette University′s Arthur J. Schmitt Scholarship Leadership Fellowship. K.D. Bassindale received training support from NICHD grant T32 HD101395 while contributing to this publication.
Eunice Kennedy Shriver National Institute of Child Health and Human Development, (grant number R15 HD093086-02, T32 HD101395).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
