Motion intent recognition for controlling prosthetic systems has long relied on machine learning algorithms. Artificial neural networks have shown great promise for solving such nonlinear classification tasks, making them a viable method for this purpose. To bring these advanced methods and algorithms beyond the confines of the laboratory and into the daily lives of prosthetic users, self-contained embedded systems are essential. However, embedded systems face constraints in size, computational power, memory footprint, and power consumption, as they must be non-intrusive and discreetly integrated into commercial prosthetic components. One promising approach to tackle these challenges is to use network quantization, which allows complying with limitations without significant loss in accuracy. Here, we compare network quantization performance for self-contained systems using TensorFlow Lite and the recently developed QKeras platform. Due to internal libraries, the use of TensorFlow Lite led to a 8 times higher flash memory usage than that of the unquantized reference network, disadvantageous for self-contained prosthetic systems. In response, we offer open-source code solutions that leverage the QKeras platform, effectively reducing flash memory requirements by 24 times compared to Tensorflow Lite. Additionally, we conducted a comprehensive comparison of state-of-the-art microcontrollers. Our results reveal that the adoption of new architectures offers substantial reductions in inference time and power consumption. These improvements pave the way for real-time decoding of motor intent using more advanced machine learning algorithms for daily life usage, possibly enabling more reliable and precise control for prosthetic users.

Sensorimotor impairment is a prevalent condition requiring effective rehabilitation strategies. This study introduces a novel wearable device for Mindful Sensorimotor Training (MiSMT) designed for sensory and motor rehabilitation. Our MiSMT device combines motor training using myoelectric pattern recognition along sensory training using two tactile displays. This device offers a comprehensive solution, integrating electromyography and haptic feedback, lacking in existing devices. The device features eight electromyography channels, a rechargeable battery, and wireless Bluetooth or Wi-Fi connectivity for seamless communication with a computer or mobile device. Its flexible material allows for adaptability to various body parts, ensuring ease of use in diverse patients. The two tactile displays, with 16 electromagnetic actuators each, provide touch and vibration sensations up to 250 Hz. In this proof-of-concept study, we show improved two-point discrimination after 5 training sessions in participants with intact limbs (p=0.047). We also demonstrated successful acquisition, processing, and decoding of myoelectric signals in offline and online evaluations. In conclusion, the MiSMT device presents a promising tool for sensorimotor rehabilitation by combining motor execution and sensory training benefits. Further studies are required to assess its effectiveness in individuals with sensorimotor impairments. Integrating mindful sensory and motor training with innovative technology can enhance rehabilitation outcomes and improve the quality of life for those with sensorimotor impairments.

The development of advanced prosthetic devices that can be seamlessly used during an individual’s daily life remains a significant challenge in the field of rehabilitation engineering. This study compares the performance of deep learning architectures to shallow networks in decoding motor intent for prosthetic control using electromyography (EMG) signals. Four neural network architectures, including a feedforward neural network with one hidden layer, a feedforward neural network with multiple hidden layers, a temporal convolutional network, and a convolutional neural network with squeeze-and-excitation operations were evaluated in real-time, human-in-the-loop experiments with able- bodied participants and an individual with an amputation. Our results demonstrate that deep learning architectures outperform shallow networks in decoding motor intent, with representation learning effectively extracting underlying motor control information from EMG signals. Furthermore, the observed performance improvements by using deep neural networks were consistent across both able-bodied and amputee participants. By employing deep neural networks instead of traditional machine learning approaches, more reliable and precise control of a prosthesis can be achieved, which has the potential to significantly enhance prosthetic functionality and improve the quality of life for individuals with amputations.

The concept of embodiment has gained widespread popularity within prosthetic research. Embodiment has been claimed to be an indicator for the efficacy of sensory feedback and control strategies. Moreover, it has even been claimed to be necessary for prosthetic acceptance, albeit unfoundedly. Despite the popularity of the term, an actual consensus on how prosthetic embodiment should be used in an experimental framework has yet to be reached. The lack of consensus is in part due to terminological ambiguity and the lack of an exact definition of prosthetic embodiment itself. In a review published parallel to this article, we summarized the definitions of embodiment used in the prosthetics literature and concluded that treating prosthetic embodiment as a combination of ownership and agency allows for embodiment to be quantified, and thus useful in translational research. Here, we reviewed the potential mechanisms that give rise to ownership and agency considering temporal, spatial, and anatomical constraints. We then use this to propose a multi-dimensional framework where prosthetic embodiment arises within a spectrum dependent on the integration of volition and multi-sensory information as demanded by the degree of interaction with the environment. This framework allows for different experiments on sensory feedback and control to be put into perspective. Considering embodiment in a spectrum tied to the interaction with the environment led us to conclude on the importance of evaluating prosthetic technologies operating in environments as close to daily life as possible.

The term embodiment has become omnipresent within prosthetics research and is often used as a metric for the progress made in technological prosthetic development, as well as a hallmark for user acceptance. However, despite frequent use of the term, the concept of prosthetic embodiment is often left undefined or used interchangeably with other distinct notions such as ownership. This terminological ambiguity complicates the comparison of studies using embodiment as a metric of success, which in turn hinders the advancement of prosthetics research. In an attempt to resolve these terminological ambiguities, we reviewed the used definitions of embodiment in the prosthetics literature, found common denominators, and developed a pragmatic perspective. We found that embodiment is often conceptualized in either of two frameworks based on body representations or phenomenology. We concluded that treating prosthetic embodiment as the combination of ownership and agency allows for measurement and thus improvement. Together with the measures and experiments to study ownership and agency suggested herein, we provide a pragmatic definition of prosthetic embodiment that allows for outcome comparisons between studies and creates a common reference for further discussions within prosthetics research.