A Novel Osseointegrated Human-Machine Gateway, Neuromuscular Electrodes, and Pattern Recognition


The use of implantable electrodes has been long thought as the solution for a more natural control of artificial limbs, as these offer access to long-term stable and physiologically appropriate sources of control, as well as the possibility to elicit appropriate sensory feedback via neurostimulation. Although these ideas have been explored since the 1960’s, the lack of a long-term stable human-machine interface has prevented the utilization of even the simplest implanted electrodes in clinically viable limb prostheses. In this thesis, a novel human-machine interface for bidirectional communication between implanted electrodes and the artificial limb was developed and clinically implemented. The long-term stability was achieved via osseointegration, which has been shown to provide stable skeletal attachment. By enhancing this technology as a communication gateway, the longest clinical implementation of prosthetic control sourced by implanted electrodes has been achieved, as well as the first in modern times. The first recipient has used it uninterruptedly in daily and professional activities for over one year. Prosthetic control was found to improve in resolution while requiring less muscular effort, as well as to be resilient to motion artifacts, limb position, and environmental conditions. In order to support this work, the literature was reviewed in search of reliable and safe neuromuscular electrodes that could be immediately used in humans. Additional work was conducted to improve the signal-to-noise ratio and increase the amount of information retrievable from extraneural recordings. Different signal processing and pattern recognition algorithms were investigated and further developed towards real-time and simultaneous prediction of limb movements. These algorithms were used to demonstrate that higher functionality could be restored by intuitive control of distal joints, and that such control remains viable over time when using epimysial electrodes. Lastly, the long-term viability of direct nerve stimulation to produce intuitive sensory feedback was also demonstrated. The possibility to permanently and reliably access implanted electrodes, thus making them viable for prosthetic control, is potentially the main contribution of this work. Furthermore, the opportunity to chronically record and stimulate the neuromuscular system offers new venues for the prediction of complex limb motions and increased understanding of somatosensory perception. Therefore, the technology developed here, combining stable attachment with permanent and reliable human-machine communication, is considered by the author as a critical step towards more functional artificial limbs