Participants

Two persons with right, unilateral, upper limb amputation participated in this study. Participant 1 (P1), a 49 year old male, had a right transradial amputation in 2009 due to traumatic injury. Participant 2 (P2), a 50 year old male, had a right transradial amputation in 2004 due to traumatic injury. Both participants were right hand dominant prior to injury. Participant 1 was implanted with 8-channel Flat Interface Nerve Electrodes (FINEs) around his median and ulnar nerves and a 4-channel Case Western Reserve University (CWRU) spiral cuff around his radial nerve in 2012. Participant 2 was implanted with 8-channel FINEs around his median and radial nerves in 2013. Subject 1’s implants were placed in the forearm, and subject 2’s implants were placed in the mid-upper arm [15]. Application of electrical stimulation pulse trains to FINEs implanted on the residual peripheral nerves results in evoked tactile and proprioceptive sensations perceived to be located on the missing hand and fingers.

Subject 1 had 13 years of education and Subject 2 had 12 years of education. Both subjects were users of myoelectric prostheses prior to enrollment in the take home study. Subject 1 had been using a myoelectric prosthesis regularly for 4 years before the take home study. Subject 2 had been using a myoelectric prosthesis regularly for 3.5 years before the take home study. Both report using a myoelectric prosthesis as their primary prosthetic device, and wear it most of the day (8–12 hours/day). The home use study took place between months 41 to 42 post-implant for both subjects.

All study devices and procedures were reviewed and approved by the U.S. Food and Drug Administration Investigational Device Exemption, the Cleveland Department of Veterans Affairs Medical Center Institutional Review Board, and the Department of the Navy Human Research Protection Program. All study procedures and experiments were performed in accordance with relevant guidelines and regulations of these institutions. Written informed consent was obtained from both subjects. The individual in this manuscript has given written informed consent (as outlined in the PLOS consent form) to publish the photograph and case details presented in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Take home sensory restoration system and study design.

A) Schematic diagram of the take home system demonstrating system components and connections. B) A participant uses the sensory restoration system to embrace a child. C) Timeline of the study. Participants used the prosthesis system without sensory feedback in stages 1 and 3 and with sensory feedback enabled in stage 2. The participants came to the lab for testing sessions before the start of the study and after each stage. D) System wear-time per stage for participants 1 and 2.

https://doi.org/10.1371/journal.pone.0211469.g001

iSens take home system

Participants wore a sensory restoration system in their homes and communities (Fig 1A and 1B). The system consisted of a myoelectric prosthetic socket, custom-sensorized Ottobock VariPlus Speed prosthetic hand, nerve stimulator, and several cables for connections between system components. The VariPlus Speed prosthetic hand is a single degree-of-freedom electric hand, which performs an open/close movement which resembles a power grasp. Participants wore their own prosthetic sockets with standard agonist/antagonist control. With this control scheme, electromyography electrodes in the prosthetic socket pick up electrical activity from the contractions of the user’s residual forearm muscles. The user contracts the forearm flexors to close the hand and contracts the forearm extensors to open the hand, where prosthesis speed is controlled by the magnitude of the muscle contraction. The augmented VariPlus Speed prosthetic hand included custom pressure sensors on the tips of the thumb, index, and middle fingers. The custom sensors consisted of a force sensitive resistor (FSR) embedded in silicone and covered with finger cots. The silicone coating conveyed forces from most common orientations of object interaction with the prosthesis and the cots protected the sensors during daily use. A custom aperture sensor that measured the width of prosthetic opening was installed underneath the prosthetic cosmetic cover. The sensor information was sent via a cable to the nerve stimulator. The stimulator transformed the information into an appropriate nerve stimulation pulse train and transmitted it through a second external cable to the participant’s percutaneous leads and then to the FINEs implanted on the residual somatosensory nerves. For both subjects, the four sensors on the prosthetic hand (3 for fingertip pressures and 1 for hand aperture) were mapped to four stimulation channels in the subject’s 8-channel FINE on their median nerve. The stimulation channels delivered biphasic, charge-balanced, cathode-first stimulation pulse trains to the median nerve. Touch on a fingertip sensor resulted in perceived sensation on the participant’s phantom hand, collocated with the position of the sensor on the prosthetic hand. For example, touch on the index sensor yielded sensation on the perceived index finger of the phantom. The aperture sensor corresponded to a tactile percept on the palm for Subject 1 and a proprioceptive sensation of hand closing for Subject 2. Changes in pressure on the fingertip sensors or changes in the width of the prosthetic aperture were translated into changes in sensation intensity using pulse frequency [18,22].

Study design

The overall home study was a quasi-experimental study that utilized a cross-over design with three stages (Fig 1C). At the start of the study, participants visited the laboratory for training on the system and in-lab testing. The first stage, designed to provide time for the participants to acclimate to wearing the system and using the new prosthetic terminal device, was two weeks in length. Participants wore the sensorized terminal device and used their normal myoelectric controls, but had the sensory stimulation turned off. During the second stage, which was intended to be a week in length, participants wore the sensory restoration system with the stimulation turned on. Finally, during the third stage, which was two weeks in length, the participants wore the same prosthesis and system with the stimulation turned off. Participants visited the laboratory at the end of each Stage for in-lab testing and activation/deactivation of the sensory feedback as appropriate.

Participant 1 wore the system in each Stage for a total of 7 days (Fig 1D). The actual length of each Stage exceeded 7 days because there were several instances of device malfunction where the system hardware was unusable for 3–4 days while it was being repaired. Participant 2 wore the system for 12, 13, and 15 days in Stages 1, 2, and 3, respectively. In Stage 2, Participant 2 wore the system with sensation for 12 consecutive days. In this case, Stage 2 was extended to accommodate the participant’s travel schedule. Stage 3 was also extended for Participant 2 due to an illness that prevented the participant from travelling to the laboratory (Fig 1D).

Each day, the participants independently donned the system and made all cabling connections. On days in which sensation was enabled, participants calibrated their stimulation settings to ensure that the perceived sensations were suprathreshold and comfortable. Participants could modify the stimulation settings whenever they chose during Stage 2.

Data collection

During the home study, participants completed daily surveys about their experiences and sensations. In laboratory testing sessions, they completed additional surveys related to perceptions of prosthesis use and performed standardized functional tasks with the prosthesis. The quantitative portions of this data have been reported previously [22].

The present manuscript analyses the qualitative data collected throughout the study, which includes participants’ written responses to free response questions on daily diaries and in-lab surveys and verbal responses to an unstructured interview with researchers in the laboratory setting. Participants completed daily diaries in which they were encouraged to write about the prosthesis, the sensations evoked by the sensory stimulation, and any other comments about their experiences (see S2 Appendix for daily diary free response questions). At the beginning of the study and after each stage, participants completed testing sessions in the laboratory environment in which they performed standardized functional tasks and described their experiences with these tasks using a written free response form (see S2 Appendix). At the conclusion of Stage 2, they participated in an unstructured interview with research staff. Interview questions focused on the sensory experience and sensory restoration system (see S2 Appendix), with additional probing by the interviewer to clarify or expand comments. Interviews were audio recorded and transcribed.

Data analysis

This qualitative study aimed to develop a complex, holistic understanding of participants’ experience with the sensory restoration system (SRS). Data analysis was performed by three investigators using a grounded theory approach with constant comparison methods [31]. The intent of grounded theory is to generate a framework explaining the inter-relationships of concepts related to a particular situation [32,33]. This approach is appropriate when studying a new phenomenon where the relationships between concepts are poorly understood [34].

Three of the investigators (EG, LR, AG) independently read the interview transcripts and written responses from the diaries and performed line by line open coding of key concepts they observed within the data [33]. NVivo 11 software was used to organize data (NVivo qualitative data analysis Software; QSR International Pty Ltd. Version 11, 2017). NVivo stores source documents in a central repository, and enables linking of segments of text to user-specified labels (nodes). NVivo enables users to group text segments by node, and can generate statistics and visualizations of coded content.

Initial codes were discussed and compared, and iteratively fine-tuned through consensus of the three data analysts. The codes were organized into several over-arching categories and further divided into sub-categories. Complete descriptions of each code were generated and supported by rich text exemplars. These descriptions were then reviewed by the three analysts, which resulted in iterative refinement in the coding categories, their descriptions, and organizational structure. Repeated refinement of the coding structure led to the final coding structure shown in Table 1. Memos of the analytic process were written to serve as an audit trail of decisions made during data analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Coding structure.

Left column indicates category, middle column indicates sub-category, and right column describes the content coded within each sub-category. White rows were included in the model; grey rows were not.

https://doi.org/10.1371/journal.pone.0211469.t001

The instances of overlap among the categories were calculated using NVivo. NVivo does this by finding segments of text that are coded into multiple categories. These text segments are then scored as an instance of overlap among the categories of which it is a member. The relationships uncovered in the coding process were discussed and used in the process of axial coding to identify the key relationships between major categories and subcategories [33]. Directionality of the relationships between categories was determined by examining the data contained in the overlap between pairs of categories. Axial coding was then discussed with the 4^th investigator (DT) who participated in the process of selective coding to define the central category that we believed represented the primary theme of the research [33]. The selective coding led to the development of a theoretical framework we named, “the impacts of sensation.” The key components of the theory were refined by re-reviewing the raw data for consistency and gaps [35]. In general, the categories excluded from the model were those that either did not interact with other categories or those that focused on the specific details and implementation of this particular sensory restoration system, and thus did not seem to reflect the broader concept of having sensory feedback.

Return to normalcy

Comments pertaining to the overall category of return to normalcy were coded in the following sub-categories: naturalness of the sensation experience, focus and visual attention required when using a prosthesis, acclimation to the sensory stimulation, sense of embodiment of the prosthesis, phantom position, and using the prosthesis in interpersonal interactions. Except for the category of phantom position, which is described in S1 Appendix, each sub-category is described below.

Self-efficacy

The category of self-efficacy included comments pertaining to confidence (or lack of confidence) in using the prosthesis to do tasks. We identified the following sub-categories of self-efficacy: perceptions of the usefulness of the sensory information provided by the system, confidence (or lack of confidence) in the system and their perceived abilities, and control interference with prosthesis use. The sub-categories of “usefulness of sensory information” and “confidence in tasks” are included in our overall model and described below; the other sub-categories are described in S1 Appendix.

Sensation experience

We coded comments pertaining to the experience of specific dimensions of the sensory percepts using the following sub-categories: modality, location, intensity, and timing. The sub-category of modality, part of our overall model, is described below; the remaining sub-categories are described in S1 Appendix.

User attitudes

Comments coded in the “user attitudes” category pertained to the participants’ views of the benefits of sensory stimulation (or lack of benefit) and the impact of the sensory stimulation system on prosthesis wear time.

The impacts of sensation on the experience of a hand

We labeled the core category generated from the analysis, “the impacts of sensation.” This category characterizes the lived experience of having a sensing, embodied, prehensile upper limb, rather than an insensate prosthetic tool. Our resulting theoretical model explains the major factors contributing to this complex process. The full model (Fig 2) incorporates key sub-categories of return to normalcy, self-efficacy, sensation experience, and user attitudes. The critical categories of return to normalcy included in the model are the perceived naturalness of the sensory experience, the experience of embodiment of the prosthesis, the focus and attention required to use the prosthesis, and the use of the prosthesis in social interactions. The model shows the relationships between these categories to two key aspects of self-efficacy, which we call usefulness of sensory information and confidence in abilities, and an aspect of user attitudes, which we call outcome acceptance. The influence of sensation modality, a key aspect of sensation experience, on these relationships is also shown. The model demonstrates the directional relationships of these categories, as determined by analysis of the text contained in the overlap between categories. The sub-category of acclimation, part of return to normalcy, is a process that strengthens or modulates the relationships between some categories over time. There are four main nodes in our model, each of which interacted strongly with several other categories. These four nodes are: naturalness, usefulness, embodiment, and outcome acceptance, and each are described in more detail below.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Model of the benefits of sensation on the experience of a hand.

Concept categories are indicated by color. Solid arrows indicate directional relationships between concepts. Dashed arrows indicate relationships that are influenced by acclimation and thus strengthen over time.

https://doi.org/10.1371/journal.pone.0211469.g002

Perceived naturalness largely results from the experience of sensation modality provided by the SRS and the usefulness of sensory information (Fig 3A). The sensory percepts become increasingly more natural feeling through the process of acclimation, which is achieved as the user garners experience using the system in everyday life. The usefulness of sensory information enables more spontaneous and fluid prosthesis use, attained, in part, as the user returns to a more normal state, with a greater sense of naturalness facilitated by reduced dependence on visual information when using the prosthesis (i.e. the category of reduced focus and attention) and a lessening of cognitive burden, both of which are also enhanced with acclimation. Thus, everyday activity performance with the SRS becomes less mentally taxing, and the experience of the sensorized prosthetic hand becomes more natural.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Concept relationships governing the benefits of sensation on the experience of a hand.

A) Determinants of naturalness of experience. B) Determinants of usefulness of sensory information. C) Determinants of embodiment. D) Determinants of outcome acceptance. Solid arrows indicate directional relationships between concepts. Dashed arrows indicate relationships that are influenced by acclimation and thus strengthen over time.

https://doi.org/10.1371/journal.pone.0211469.g003

The perceived naturalness of sensory experience also contributes directly to increasing the perceived usefulness of the sensory information, because information from more natural feeling percepts is more easily utilized (Fig 3B). Certain sensory modalities are particularly valuable in enhancing the usefulness of the supplied sensory information; sensory information is more useful when it is conveyed via a modality that matches the physical nature of the information (for example, prosthetic aperture information is more useful when the associated evoked sensation feels like joint movements rather than tingling or some other tactile modality).

Prosthesis embodiment, or the experience that the prosthesis is a part of oneself, is created directly through the perceived naturalness of the sensory experience, through the modality of the sensory percepts, and through a greater sense of confidence in abilities (Fig 3C). Confidence is enabled by the usefulness of sensory information in everyday life. The naturalness of the experience of having a hand that can feel transforms the prosthesis from an extracorporeal tool into a normal, animate hand. Having sensory information that was useful in performing tasks yields increased confidence in their abilities to do tasks with the prosthesis. Higher confidence supports the perspective that the prosthesis is a capable and rightful part of the body. The reduced focus and visual attention required to operate the prosthesis, facilitated by the usefulness of the sensory information, further strengthens the perception that the prosthesis is a natural part of the body.

The model culminates in the category of outcome acceptance, which is characterized by satisfaction with the SRS and a desire to have and use it (Fig 3D). Outcome acceptance is driven by one’s confidence in abilities, the sense of prosthesis embodiment, the overall gratification created from restoration of hand sensation modalities, and enhanced interaction with others. The usefulness of sensory information indirectly leads to outcome acceptance, since it can enable greater confidence in abilities, which also impacts one’s interactions with others. The sensation experience, and in particular sensation modalities (even those considered less than ideal), was integral to the participants’ attitudes and acceptance of their situations. The naturalness of the experience of having a hand that can feel, through its contribution to embodiment, also contributes to outcome acceptance.

Understanding the perceived benefits of sensory restoration

Our analysis produced a theoretical model that describes the perceived benefits of sensation on the overall experience of having a hand. This is the first study to examine the personal experiences, opinions, and attitudes of upper limb prosthesis users who used prosthetic hands with multi-location, naturalistic, somatotopically-matched sensory feedback. We believe our model will be useful for understanding the impact of restoring sensation to upper limb amputees, especially through peripheral nerve interfaces.

Major components of our model are supported by the quantitative data collected during the take home study and presented in a parallel publication [22]. During the home study, participants completed daily surveys about their experiences and sensations. In laboratory testing sessions, they completed additional surveys related to perceptions of prosthesis use and performed standardized functional tasks with the prosthesis. Surveys included the Orthotics Prosthetics Users’ Survey (OPUS) Upper Extremity Functional Status (UEFS), which measures the perceived difficulty of doing activities of daily living with the prosthesis; the Patient Experience Measure (PEM), which measures perceptions of self-efficacy (confidence in abilities), prosthesis embodiment, social interaction, body image, and prosthesis efficiency; OPUS Quality of Life (QoL) survey, which measures quality of life; and the Quick Disabilities of the Arm, Shoulder, and Hand (QuickDASH), which measures perceived disability. Throughout the study, on-board data logs recorded wear time and prosthesis sensor activity [22].

Survey data from the PEM and OPUS UEFS demonstrated that after using the sensory restoration system at home for a week, participants had significant improvements in self-efficacy, prosthesis embodiment, and perceptions of their abilities to use the prosthesis in social interactions. This supports the key impacts of sensation, elucidated in the theoretical model, on the categories of confidence in abilities, embodiment, and interaction with others. Prosthesis sensor data indicated that participants wore their prostheses longer and used their prostheses more frequently to grasp or interact with objects when sensation was enabled, substantiating the participants’ statements about the impact of sensation on wear time. Finally, participants’ perceived disability scores on the QuickDASH were significantly decreased and quality of life scores on the OPUS QoL were significantly increased after home use with sensation, which quantitatively demonstrate the impact of sensation on aspects of outcome acceptance [22].

We believe that our theoretical model helps frame and interpret findings from prior studies of sensory restoration. Our model can also be used to facilitate the synthesis of results across previous and future sensory restoration studies by grouping common concepts and describing interactions among experimental variables. The model may be applied to identify themes from larger data sets that may reflect higher-order changes in perceptual or cognitive processes due to sensory restoration. Additionally, the model provides a framework that can be used to inform the development of novel instruments that measure important constructs related to sensory restoration.

Our study also uncovered participant views on which technological aspects of the sensory feedback system need to be improved in future design iterations. Robustness was a concern for both participants, and both avoided using the device to do outdoor activities that may have damaged the system (see “Lack of confidence/distrust of system” in S1 Appendix). Participants experienced instances where the sensory stimulation interfered with their myoelectric control (see “Control interference” in S1 Appendix), suggesting that future engineering efforts should address strategies to reduce stimulation artifact in EMG recordings. There were also several issues that participants identified with the software of the stimulator and the cabling of the system components (see “System Operations” category in S1 Appendix). For example, the participants were able to independently calibrate their stimulation levels as desired each day, but did not like that the levels reset to predefined settings when the stimulator was power cycled. Researchers and engineers should incorporate this user feedback into their designs for future sensory restoration systems in order to improve the user experience, which may also influence outcomes.

Comparison to upper limb amputee outcome acceptance

In keeping with grounded theory methodology, we were largely naïve to the qualitative literature on models of outcome acceptance in upper limb amputees during our data analyses. After completing our analyses, we conducted a post-hoc literature search on the topic. We found remarkable similarities between our model and other models presented in the literature on upper limb prosthesis use.

For example, key components of our model of outcome acceptance mirrored the major themes identified in Murray’s phenomenological study of successful prosthesis users, despite the fact that our participants had sensory restoration whereas those in Murray’s study did not [27]. Our sub-categories of “acclimation” and “focus and visual attention” parallel the themes “adjusting to a prosthetic” and “awareness of the prosthesis,” respectively. In “adjusting to a prosthetic,” Murray discussed the process through which a user becomes familiar with the prosthesis, both physically and perceptually. In “awareness of the prosthesis,” Murray described how users experienced a gradual decrease in the amount of attention and concentration required in prosthesis use, and an increasing “naturalness” of prosthesis use with experience. Similarly, we saw a direct relationship between “focus and visual attention” and “naturalness of experience.”

Comparison of our model to themes identified in prior studies of new prosthesis technologies also revealed several parallels. A study by Luchetti and colleagues examined outcome acceptance for persons who used a Michelangelo prosthetic hand for six months [26]. In Luchetti’s study, the participants described their myoelectric device as “essential” to their lives, and expressed feeling useful with the device donned [26]. Our participants also expressed a similar sentiment of wanting to wear the sensorized prosthesis for longer periods and to use it more frequently to grab objects, which contributed to our theme of “outcome acceptance.” As in our study, Luchetti et al. found that the use of a prosthesis to engage in social interactions, such as shaking someone’s hand, was important to participants and linked to outcome acceptance. However, the motivation of our subjects for engaging in social interactions with the sensorized prosthesis was driven by the prosthesis feeling natural (naturalness of experience) and like their hand (embodiment) due to the sensory restoration. In contrast, Luchetti and colleagues surmised that the social interactions of their participants were encouraged by the “normal” appearance and functionality of the Michelangelo Hand [26].

A phenomenological study by Lundberg and colleagues examined the experience of living with an osseointegrated prosthesis [11]. Lundberg et al found that the enhanced functional capabilities enabled by their technology allowed participants to be “more engaged in their life and social interactions” [11], which aligns with our finding that sensation led to improved “social interactions.” Lundberg and colleagues reported that their participants stated that osseointegration allowed them to “not need to spend time and mental energy” on their prostheses, which the authors posited contributed to the functional changes enabled by osseointegration. This finding aligns well with our “reduced focus and attention” category. Just as our participants experienced “distrust of the system” in some cases (see S1 Appendix), the participants with osseointegration felt “caution” in regards to their devices and “fear of losing the improved function.” For both the sensory restoration and osseointegration studies, the participants were early adopters working with new prototypes and imperfect devices. These feelings of caution and distrust may reflect the participants’ reasonable acknowledgement of the risks they undertook while living with the intervention, or they may reflect the strong preference for having the new intervention compared to not having it due to malfunction or breakage.

Finally, Lundberg and colleagues posited that osseointegration enabled “a gradual change in their identity from considering oneself as being disabled to having an identity as a healthy person.” This concept is parallel to our domain of “return to normalcy,” and in particular the sub-categories of “acclimation” and “naturalness of the experience.” In our study, the participants’ comments demonstrated that they felt that sensation began to restore them to a more “normal” state, that is, the state of an able-bodied person. We divided this “normalcy” concept into several sub-categories, one of which was the “naturalness of the experience,” which described that the sensation itself felt “natural”, that the sensation arising from touch on the prosthetic hand seemed like a “normal” way to get information, and that the sensation enabled the overall experience of the hand to be more “normal.” We defined the category of “acclimation” as a gradual process that enabled improved naturalness of the experience over time. In both osseointegration and sensory restoration, perhaps the gradual rate of these changes is essential to the profundity of the impact of the intervention.

Implications for embodiment

In the literature, discussions of embodiment encompass the related, but differentiable concepts of body ownership, body image, and body schema. Body ownership, also called self-attribution [36] or self-identification [37], is the conscious experience that the body belongs to oneself and the identification of one’s body as distinct and perceptually unique from other people’s bodies and external objects [37–39]. Body image refers to the conscious perception of one’s body, including perceptual experiences, beliefs, and attitudes about the body [40]. In contrast, the body schema is a preconscious internal model of the body, which consists of neural representations of motor functions and multimodal sensory inputs [40,41]. Incorporation of a device into the body schema involves changes to brain representations, changes in sensory processing, and inclusion of the device in preconscious sensorimotor models [41–45]. However, a conceptual distinction is often made between true bodily incorporation, in which an object becomes part of the body, and bodily extension, in which an object is an extension of the body. [42,46]. True incorporation of non-corporeal objects into the body is thought to only be possible where the object substitutes for a missing body part and thus contributes to body completion or wholeness, such as with prostheses for persons with limb loss [42]. Bodily extension, in contrast, occurs with most forms of skilled tool use. Although bodily extension also involves changes to sensorimotor models, it is temporary and depends upon the location of the object relative to the body [42,46–48].

Our category of embodiment primarily consisted of the concepts of body ownership [42] and changes in the conscious perception of the body image [40]. Participants’ comments referring to the prosthesis as “my hand” or “my fingers” demonstrate self-attribution and a sense of ownership. Participant statements that “it felt like my hand was there” or other descriptions of their hand being continuously present speak to a body image that now includes a hand. The participants frequently contrasted the experience of having a “hand,” when sensation was enabled, with having a “tool” or “attachment,” when sensation was disabled. This also reflects a perceptual shift towards body ownership and inclusion of the prosthesis into the body image when sensation is provided.

Participants did not directly discuss body schema in this study, which is expected since the body schema operates largely outside of conscious awareness [42]. However, indirect evidence suggests that the participants likely experienced true bodily incorporation of the prosthesis. Some of the participant experiences categorized in the “reduced focus and attention”, “naturalness of experience,” and “confidence in abilities” categories likely represent consciously perceivable consequences of changes in the body schema due to incorporation of the prosthesis [40]. With the stimulation-evoked sensory feedback included in sensorimotor models of hand interactions, participants were able to grab objects with less focus (see “Reduced focus and attention”). This speaks to the “transparency” in tool experience that occurs for incorporated devices, where the tool enables an experience but is not itself the focus of the experience [42,43]. As the participants gained experience with the system and the sensory feedback became more natural, the participants began viewing the prosthesis sensation as a “normal function,” and were able to incorporate it into their decision-making processes (see “Naturalness of experience”). Further, even though we did not modify the control scheme of the prosthesis in this study, participants claimed that they had “better control” of their prosthesis and that they had “more control with less thought” (see “Confidence in abilities”). These comments suggest that the sensation was able to close the loop in their internal sensorimotor control models, at least in part, reflecting the successful incorporation of the sensory-enabled prosthesis into the body schema.

The critical role of embodiment

Our model demonstrates that embodiment is critical for outcome acceptance. The pivotal role of prosthesis embodiment in prosthetic outcome acceptance has been acknowledged by others who have written about the importance of limb ownership, body image, and prosthesis incorporation into the body [27,37,47]. Prior studies describe prosthesis users’ desire for a “corporeal” rather than an “extracorporeal structure” or “inert supplement” to the body. This speaks to prosthesis users’ preference for a device that becomes a part of one’s own body through bodily incorporation, rather than an extension of the body schema or a tool [49]. The importance of embodiment also emerged from prior qualitative and phenomenological studies of successful prosthesis users [27], users of osseointegrated prostheses [11], users of dexterous prostheses [26], and the experience of living with limb loss [29]. Body image has been noted to contribute to psychosocial adjustment to amputation, especially in women [25,50]. In addition, a Delphi study of rehabilitation professionals and upper limb amputees identified several factors related to body image and incorporation as indicators of successful prosthesis outcomes and important factors to consider in rehabilitation [51].

Our model shows a direct relationship between sensation modality and embodiment, supporting the assertion that a primary condition for prosthesis incorporation into the body is that the prosthesis becomes a “knowing body part” [42,43]. The “knowing” provided by the sensory stimulation in our study clearly contributed to the sense of embodiment. This is supported by prior studies which found that touch and proprioceptive signals are important for establishing body ownership in the rubber hand illusion [36,52–54]. The relationship between sensation and embodiment was demonstrated in a prior qualitative study of upper limb prosthesis users, which included those who had experienced vibrotactile or mechanotactile sensory substitution on the residual limb. These subjects reported that having sensation is important to establishing prosthesis incorporation [30]. Another qualitative study of the experience of osseointegrated prosthetic limbs focused on the varying degrees to which participants incorporated their prostheses into their sense of self, further highlighting the importance of prosthetic incorporation to amputees [11]. Sensation likely played a role in the incorporation of the prosthesis for these participants, given that participants reported improved proprioception, gained an ability to feel through the skin of the residual limb that was no longer encased in a socket [11], and likely experienced bone-conveyed sensation via osseoperception [12].

The relationship between sensation and embodiment has also been demonstrated in prior studies of amputees with targeted sensory reinnervation [55], sensory feedback provided through electrical stimulation [21,56,57], and proprioception provided through surgical coupling of agonist and antagonist muscles [58]. Interestingly, with targeted sensory reinnervation, proprioceptive sensation alone did not provide a sense of embodiment of a prosthesis [59], while cutaneous touch feedback was able to elicit a sense of embodiment [55]. Our subjects experienced both tactile and proprioceptive sensations during the home use study. The interaction of sensation modality with embodiment in our theoretical model may reflect that these modalities contributed to prosthesis embodiment to different extents.

Although our study and prior studies found that sensation played a critical role in prosthetic embodiment, sensory experience may not be an absolute requirement for embodiment. Murray described how some, but not all, successful prosthesis users experienced the prosthetic limb as a corporeal structure (part of themselves), even though these users did not have sensory feedback [27]. Luchetti and colleagues also observed the theme of embodiment among their participants who utilized a novel prosthetic hand without sensation, and attributed it to the hand’s natural appearance and enhanced dexterity [26]. Body congruence, i.e. that objects resemble body-parts, is considered by some to be a necessary element of successful object incorporation [36,60]. Training and learning with a prosthesis may also enable increased prosthesis incorporation into the body schema, even without sensation [61,62].

However, these disparate findings regarding the role of sensation in embodiment may be resolved by considering that there may be varying degrees of embodiment and that the likelihood of embodiment also differs across persons. De Vignemont describes a “manifold” of various types of embodiment, proposing that there are differing degrees of embodiment [47]. This idea of varying degrees of embodiment is supported by studies of anatomical congruence of embodied objects [60] and the conceptual distinction between true bodily incorporation and bodily extension described above [42]. In addition, not all prosthesis users currently experience embodiment [27]. While both participants in our study provided strong evidence that they experienced limb ownership and bodily incorporation of their prostheses, approximately half of the participants in a prior qualitative study of osseointegration did not experience full incorporation [11].Thus, although prosthesis functionality, appearance, and training may contribute to prosthetic embodiment, they are not always sufficient. The addition of sensory restoration to prostheses could lead to greater likelihood of prosthetic embodiment in upper limb amputees and enhance the degree of embodiment for those who experience it.

Implications for phantom pain

Phantom limb pain is another aspect of the experience of limb loss that has been well-documented and may be a factor related to outcome acceptance [63]. Because our participants did not experience phantom limb pain after the start of the study [15], we could not study the impact of sensation on phantom limb pain or the interaction of phantom limb pain with other concepts in the model. However, based on evidence from prior studies of electrical stimulation of peripheral nerves, sensory restoration may be an effective treatment for phantom pain [15,64–66]. We hypothesize that daily exposure to sensory restoration through at-home usage of a sensory-enabled prosthesis will yield further reductions in phantom pain beyond those achievable in laboratory settings. Interestingly, limb telescoping, which is the experience of the phantom limb location retracting into the stump [67,68], was significantly decreased for our participants when sensation was enabled during the home study, such that the location of the phantom limb attained a more natural extended position when they were provided sensory feedback [22]. Limb telescoping is correlated with phantom limb pain, and although the neural mechanisms are not fully understood, both telescoping and phantom pain are related to cortical reorganization in the somatosensory cortex [68,69]. Sensory feedback may reverse the cortical reorganization of phantom limb areas, leading to decreased phantom pain and improved body representation [70]. In our home study, when sensory feedback was again disabled in Stage 3, the telescoping experience returned and the phantom limb retracted towards the stump [22]. This suggests that continued exposure to behaviorally-relevant sensory feedback may be necessary to maintain the cortical organization changes induced by sensation.

Limitations

Our model was developed from data derived from the lived experiences of two male research participants with upper limb amputation who had a sensory restoration prosthesis system. Both research participants had the same neural interface for sensory restoration, the same sensory restoration system, and the same prosthetic hardware. Both had transradial amputation due to trauma. Therefore, we cannot tell to what extent our model will generalize to other sensory restoration systems, neural interfaces, terminal device types, to women amputees, or to other patient populations.

We acknowledge that the model may be incomplete and limited because of certain aspects of our data collection methodology. Data collected via free response survey questions may have been limited by the ability of the participants to write their answers with their non-dominant hand or prosthesis, given that both lost their dominant hand due to amputation. In addition, daily diary responses may have resulted in over-representation of certain concepts, since participants may have chosen to write about similar themes repeatedly across days. That said, the majority of the data we analyzed in this study was accrued through participant verbal responses to unstructured interviews. Thus, the model may overemphasize concepts uncovered through interview. The unstructured interviews were collected by several interviewers, each of whom had their own style of question phrasing and probed responses differently. It is possible that a more structured interview approach with standardized probing and questioning would have yielded further findings across participants. Additional research, using a larger patient population and a more structured approach, is recommended to explore and confirm this model.

Another limitation is that we did not collect any quantitative metrics of user attitudes or acceptance of the sensory-enabled prosthesis. The theoretical model that we present clearly suggests that surveys measuring user attitudes or prosthesis acceptance should be collected in future home use studies of sensory restoration prostheses. Our study only examined the impact of sensation on outcome acceptance. Thus, the interaction of sensation with other factors that have been previously identified to be important to outcome acceptance, such as hand dexterity and appearance [26], cannot be determined. Because our participants used the same single degree-of-freedom prosthesis with the same control scheme in this study, we cannot tell to what extent hand dexterity or control scheme impacted sensation experience or outcome acceptance.

Our model may also have been impacted by the perceived reactions of family and friends to the sensory restoration system. Family and social support has been shown to be critical for psychosocial adjustment to amputation [25,50,71,72]. Not surprisingly, given that this was the first study of a sensory restoration prosthesis, both participants described the skepticism of their family and friends in the functionality of the sensation. It is possible that the skepticism of the participants’ family and friends influenced their perspectives on sensory feedback, their views of the sensory-enabled prosthesis, and their choices in how to respond to survey and interview questions. Future studies of sensory restoration prostheses may yield different outcomes or demonstrate differences in our model once sensory restoration becomes more widely-known and accepted by the public.

Application of the model

The experience of having a hand is a complex phenomenon. In this study, participants had an experiential shift in which their prosthesis moved closer to being perceived as a normal hand due to the addition of sensation. The theoretical model of transformation of hand experience may serve as a starting point for further improvements to rehabilitation following limb loss.

This model could guide researchers, clinicians, or engineers in determining how best to improve outcome acceptance using sensory restoration, other novel prosthesis designs, or therapeutic interventions. For example, from the perspective of an engineer, efforts could be made to enhance outcome acceptance by designing a new sensory encoding algorithm (to increase the “usefulness of sensory information”) and/or by making the user interface easier and more intuitive to use (to increase the “naturalness of the experience”). Such efforts can then be evaluated by examining end-users’ “confidence in abilities” or prosthesis “embodiment.” From the perspective of a rehabilitation therapist, efforts could be made to improve training on use of the sensory information (to increase the “usefulness of the sensory information”) and then studied by measuring the impact of such training on “confidence in abilities” and “interaction with others.”

The similarities between our model and prior models may reflect over-arching viewpoints and needs of the amputee community and suggest a common theoretical framework to describe outcome acceptance in upper limb prosthesis users more broadly. These findings support the potential transferability of our model to all upper limb prosthesis users, and its application to aid in understanding the impact of novel prosthesis technologies and rehabilitation interventions on outcome acceptance. We recommend that the relationships outlined in our model be explored and tested in other studies of upper limb amputees to determine if the model is applicable and useful.

Though this theoretical framework was developed to explain the experience of upper limb amputees, it may be applicable to individuals with various conditions involving sensory deficits of the hand. For example, following a stroke, nerve injury, or spinal cord injury, a person’s perception of their hand or limbs may be substantially altered. One qualitative study of the experience of persons with neonatal brachial plexus injury identified several key topics that overlap with domains in our model, such as sensation experience, arm/hand compensation and preference, explaining functionality/appearance to others, self-esteem, and body image [73]. This provides evidence that there are universal aspects of the experience of living without sensation in one’s hand that are independent of etiology. Our model may also be useful for understanding the experiences of persons who use other types of assistive devices, such as wheelchairs [74]. The experience of human-device integration, which we examine here through sensory-enabled hand prostheses, likely shares common themes across devices and patient populations.