Take home sensory restoration system. (A) Participants wore their own prosthetic socket with standard agonist/antagonist myoelectric control. They were provided a VariPlus Speed™ prosthetic hand (Ottobock, Vienna, Austria), which had been augmented with Flexiforce™ pressure sensors (Tekscan, Inc., Boston, MA) embedded in silicone in the pads of the thumb, index, and middle fingers. A custom aperture sensor underneath the cosmetic glove encoded the position of the prosthetic hand’s single degree of freedom. The sensor information was sent through a cable to the external nerve stimulator, which converted the sensor information into electrical stimulation pulses. The stimulation traveled to the nerve via percutaneous leads to Flat Interface Nerve Electrodes (FINEs) implanted on the participants’ median nerves. Illustration courtesy of Cleveland FES Center. (B) Image of a subject wearing the sensory restoration system. (C) Study timeline. The study was a three-stage crossover design, in which the subjects used the system at home either without sensation (stages 1 and 3) or with sensation (stage 2). Functional metrics were administered in laboratory testing sessions at the start of the study and after each stage. (D) Study stage durations for subject 1 (left) and subject 2 (right). Subject 2 wore the system for nearly twice as many days per stage as subject 1. Subject 1 experienced 3 days of interruption per stage due to component breakage (see Methods). (E) Locations of sensory percepts associated with each prosthetic hand sensor. Sensation locations reported daily throughout the study were overlaid such that regions of higher opacity were more-frequently reported. (F) Stimulation charge delivered to each channel for subject 2 over the course of the sensory-enabled stage of the study. Participants could calibrate stimulation settings (pulse amplitude and pulse width) whenever they chose. Filled dots are stimulation settings at each recalibration; color corresponds to the percept locations in subpanel E. The slopes of the regressions for the thumb, index, and middle channels (navy blue, teal, and magenta) were not significantly different from zero (regression slope test, p > 0.1 for all). The slope for the aperture channel (brown) was −0.11 nC/hr and was significantly less than zero (p = 0.002).