PUBLISHED ON IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, MAY 2017 1 Embedded System for Prosthetic Control Using Implanted Neuromuscular Interfaces Accessed Via an Osseointegrated Implant Enzo Mastinu, Student Member, IEEE, Pascal Doguet, Member, IEEE, Yohan Botquin, Bo H˚ akansson, and Max Ortiz-Catalan, Member, IEEE, Abstract—Despite the technological progress in robotics achieved in the last decades, prosthetic limbs still lack functional- ity, reliability, and comfort. Recently, an implanted neuromuscu- loskeletal interface built upon osseointegration was developed and tested in humans, namely the Osseointegrated Human-Machine Gateway. Here we present an embedded system to exploit the advantages of this technology. Our Artificial Limb Controller allows for bioelectric signals acquisition, processing, decoding of motor intent, prosthetic control, and sensory feedback. It includes a neurostimulator to provide direct neural feedback based on sensory information. The system was validated using real-time tasks characterization, power consumption evaluation, and myoelectric pattern recognition performance. Functionality was proven in a first pilot patient from whom results of daily usage were obtained. The system was designed to be reliably used in activities of daily living, as well as a research platform to monitor prosthesis usage and training, machine learning based control algorithms, and neural stimulation paradigms. Index Terms—Electromyography (EMG), prosthetic controller, osseointegration, pattern recognition, Osseointegrated Human- Machine Gateway (OHMG), sensory feedback. ACRONYMS AFE Analog Front-End ALC Artificial Limb Controller DC Direct Control EMG Electromyography ENG Electroneurography LDA Linear Discriminant Analysis MCU Microcontroller Unit MPR Myoelectric Pattern Recognition MSPU Mixed Signal Processing Unit NS Neurostimulator OHMG Osseointegrated Human-Machine Gateway PCCU Prosthetic Control and Communication Unit SVM Support Vector Machine Research supported by Swedish Research Council (Vetenskapsrdet), VIN- NOVA and partially from EU project DeTOP. E. Mastinu and B. H˚ akansson are with the Dept. of Electrical Engi- neering, Chalmers University of Technology, Gothenburg, Sweden (e-mail: enzo@chalmers.se, boh@chalmers.se). P. Doguet and Yohan Botquin are with Synergia Medical, Mont- Saint Guibert, Belgium (e-mail: pascal.doguet@synergiam.com, yohan.botquin@synergiam.com). M. Ortiz-Catalan is with the Dept. of Electrical Engineering, Chalmers University of Technology, and Integrum AB, Gothenburg, Sweden (e-mail: maxo@chalmers.se). I. I NTRODUCTION D ESPITE the advances in prosthetic hardware that allow an increasing number of artificial joints to approach those of the lost limb [1], a major issue remains unsolved, namely, how to achieve a reliable and natural control of the prosthetic limb. After many years of research and development on prosthetics, amputees mostly rely on direct control (DC) (also known as one-for-one control, or one-muscle to one-function), which is often combined with sequential solutions for grasp switching based on encoding unnatural muscular activation (e.g., co-contraction) [2]. This control mechanism is pervasive owing to its simplicity, relative reliability, and ease to learn. Unfortunately, the functional outcome is commonly related to the specific patient predisposition, thus often resulting in rejection of the myoelectric prosthesis, or in reduction of the robotic potential to a simple prosthesis claw [3]. Modern prostheses are hindered by discomfort and poor functionality [4–6]. The latter has pushed research towards the challenge of using information from the neuromuscular system in more surgically and technologically sophisticated manners [7–9]. Comfort and functionality had been considerably improved by the use of osseointegration for direct skeletal attachment of limb prostheses [10], [11]. Osseointegration provides a long- term, mechanically stable interface between biology and the artificial limb, in which a titanium implant is surgically in- serted into the remaining bone of the amputated extremity [11]. The osseointegration technology has been recently enhanced to allow bidirectional communication between implanted neu- romuscular electrodes and the artificial limb [12]. This Os- seointegrated Human-Machine Gateway (OHMG) combines the benefits of skeletal attachment with the reliability and increased information provided by implanted muscular elec- trodes [13], [14]. The first attempts on using implanted electrodes to restore sensory feedback were conducted over 40 years ago [15], and several others were reported more recently [16–19]. Despite the efforts, close-loop control has not been achieved yet in activities of the daily living arguably due to the lack of a suitable communication interface. The OHMG now provides a clinically viable long-term access to implanted neural interfaces that can be used for bidirectional communication to accomplish such purpose. Here we present the development of an embedded system to exploit the advantages of the OHMG technology. Efforts were placed in hardware and software design, carefully seeking a