| dc.description.abstracteng | Current commercially available prosthetic systems still lack sensory feedback and amputees are forced to maintain eye-contact with the prosthesis when interacting with their environment. Electrocutaneous stimulation is a promising approach to convey sensory feedback via the skin. However, when discussed in the context of prosthetic applications, it is often refused due to its supposed incompatibility with
myocontrol.
This dissertation now addresses electrocutaneous stimulation as means to provide sensory feedback to prosthesis users, and its implications on myoelectric control, possible use for improved or accelerated mastering of prosthesis control through closing of the control loop, as well as its potential in aiding in the embodiment of prosthetic components. First, a comparison of different paradigms for encoding sensory feedback variables in electrocutaneous stimulation patterns was done. For
this, subject ability to employ spatially and intensity-coded electrocutaneous feedback in a simulated closed-loop control task was evaluated. The task was to stabilise an invisible virtual inverted pendulum under ideal feedforward control conditions (joystick). Pendulum inclination was either presented spatially (12 stimulation sites), encoded by stimulation strength (≧ 2 stimulation sites), or a combination of the two. The tests indicated that spatial encoding was perceived as more intuitive, but intensity encoding yielded better performance and lower energy expenditure.
The second study investigated the detrimental influence of stimulation artefacts on myoelectric control of prostheses for a wide range of stimulation parameters and two prosthesis control approaches (pattern recognition of eight motion primitives, direct proportional control). Artefact blanking is introduced and discussed as a practical approach to handle stimulation artefacts and restore control performance back to the baseline. This was shown with virtual and applied artefact blanking (pattern recognition on six electromyographic channels), as well as in a practical task-related test with a real prosthesis (proportional control).
The information transfer of sensory feedback necessary to master a routine grasping task using electromyographic control of a prosthesis was investigated in another study. Subjects controlled a real prosthesis to repeatedly grasp a dummy object, which implemented two different objects with previously unknown slip and fragility properties. Three feedback conditions (basic feedback on grasp success, visual grasp force feedback, tactile grasp force feedback) were compared with regard to their influence on subjects’ task performance and variability in exerted grasp force. It was found that online force feedback via a visual or tactile channel did not add significant advantages, and that basic feedback was sufficient and was employed by subjects to improve both performance and force variability with time. Importantly, there was no adverse effect of the additional feedback, either. This has important implications for other non-functional applications of sensory feedback, such as facilitation of embodiment of prosthetic devices.
The final study investigated the impact of electrocutaneous stimulation on embodiment of an artificial limb. For this purpose, a sensor finger was employed in a rubber-hand-illusion-like experiment. Two independent groups (test, control), were compared with regard to two objective measures of embodiment: proprioceptive drift, and change in skin temperature. Though proprioceptive drift measures did not reveal differences between conditions, they indicated trends generally associated to a successful illusion. Additionally, significant changes in skin temperature between test and control group indicated that embodiment of the artificial digit could be induced by providing sensory substitution feedback on the forearm.
In conclusion, it has been shown that humans can employ electrocutaneous stimulation feedback in challenging closed-loop control tasks. It was found that transition from simple intuitive encodings (spatial) to those providing better resolution (intensity) further improves feedback exploitation. Blanking and segmentation approaches facilitate simultaneous application of electrocutaneous stimulation and electromyographic control of prostheses, using both pattern recognition and classic proportional approaches. While it was found that force feedback may not aid in the mastering of routine grasping, the presence of the feedback was also found to not impede the user performance. This is an important implication for the application of feedback for non-functional purposes, such as facilitation of embodiment. Regarding this, it was shown that providing sensory feedback via electrocutaneous stimulation did indeed promote embodiment of an artificial finger, even if the feedback was applied to the forearm.
Based on the results of this work, the next step should be integration of sensory feedback into commercial devices, so that all amputees can benefit from its advantages. Electrocutaneous stimulation has been shown to be an ideal means for realising this. Hitherto existing concerns about the compatibility of electrocutaneous stimulation and myocontrol could be resolved by presenting appropriate methods to deal with stimulation artefacts. | de |