Journal of NeuroEngineering and Rehabilitation

Background Major upper-limb amputation affected over 40,000 Americans as of 2005 [ 1 ] , with over 11,000 additional wrist disarticulation or higher-level amputations between 2005 and 2013 [ 2 ] . The impairment of bimanual dexterity caused by amputation interferes with basic activities of daily living; routine activities such as driving, household work, and leisure activities; and limits employment oppor- tunities. Currently, the most effective treatment is use of a prosthesis, and recent advances in prosthetic technology, including myoelectric devices with multi-articulating hands, pattern recognition – based control systems, and surgical techniques such as targeted muscle reinnervation (TMR) [ 3 ] have been developed to improve prosthetic function. However, a functional performance remains challenging. As prosthetic limbs and control systems become more advanced and costly it is important to quantify perform- ance benefits that these technology improvements pro- vide to users. Outcome measures may also be useful to help track progress through rehabilitation protocols and indicate specific areas which required additional therapy. The Academy of Prosthetics and Orthotics Upper Limb Prosthetics Outcome Measure committee provided recommendations for measuring functional effective- ness of prosthetic treatment/occupational therapy [ 4 ] . Rather than relying on a single test across all patients and phases of device development, they recommended using multiple test formats to capture all aspects of performance. Promising tests included the Assessment for Capacity of Myoelectric Control (ACMC), the Southampton Hand Assessment Procedure (SHAP), a modified Box and Blocks test, the Jebsen-Taylor Test, and a Clothespin Relocation task. Of these tests, only the ACMC has been validated and demonstrated to have good test-retest reliability for the field of upper-limb prosthetics [ 5 ] . The remaining tests have been identified as promising tests to use, particularly when performing research and development studies in the field of upper-limb prosthetics [ 6 ] . An alternative method of assessing performance is to use virtual environments or serious gaming — video games or virtual environments designed for training purposes. Proponents of these tools promote their economic benefits, their manageable and rapid devel- opment, and the availability of powerful computing and processing technologies as factors driving the re- cent success and popularity of simulated environ- ments for clinical and research applications. Virtual tools have been developed for stroke rehabilitation [ 7 ] , assessment of children with cerebral palsy [ 8 ] , and for other neuromuscular disorders [ 9 ] . Several virtual environments have been proposed for myoelec- tric control applications. Virtual environments for myoelectric control have evolved from rudimentary graphical user interfaces to more life-like virtual avatars and real-time practice envi- ronments, with performance tasks such as virtual clothespin movement tasks [ 10 ] , posture matching tasks [ 11 ] , or Fitts-law style target acquisition tasks. Alterna- tive approaches have abstracted the experience away from controlling a prosthetic limb, instead using myo- electric signals as inputs to engage in commercially available video games, such as Guitar Hero ™ [ 12 ] or custom-designed serious games [ 13 ] . Intuitively, one would expect improved performance or testing scores within a virtual environment to trans- late into better prosthesis control, which would in turn lead to better functional outcomes. However, this as- sumption has not been thoroughly tested. Powell et al. [ 14 ] showed that upper limb prosthesis users could con- trol a virtual prosthesis better after practicing pattern recognition control in a virtual environment across mul- tiple days; however, functional tests with a physical pros- thesis were not reported. Recently van Dijk et al. [ 13 ] demonstrated transfer of myoelectric control skills after serious gaming, but only if the game was designed to en- courage behaviors specific to controlling a prosthesis. In addition, van Dijk ’ s study was limited in that it was per- formed with able-bodied subjects. The primary objective of this study was to determine the relationship between performance on a virtual test — the Target Achievement Control (TAC) test — and performance with a physical prosthesis. The secondary objective was to determine whether, after extensive oc- cupational therapy, allowing subjects to practice with a pattern recognition – controlled prosthetic arm during a 6-week home trial would further improve functional outcomes. Methods Nine individuals with transhumeral level amputations who had previously undergone targeted muscle reinner- vation (TMR) surgery were recruited for the study (Table 1 ) . All subjects, with the exception of P9, were myoelectric prosthesis users prior to enrolling into the study but not all routinely used their prosthesis. TMR has previously been described in detail [ 3 ] . In this surgical procedure, severed motor nerves that previ- ously controlled arm and hand function, are transferred onto denervated target muscles — muscles that no longer serve a biomechanical function after amputation. After reinnervation, target muscles serve as biological ampli- fiers of the motor control commands intended for the missing arm and thus provide physiologically appropri- ate EMG control signals, making prosthesis control in- tuitive. For example, after reinnervation of a segment of biceps muscle by the transferred median nerve, Hargrove et al. Journal of NeuroEngineering and Rehabilitation 2018, 15 (Suppl 1):60 Page 22 of 72