Suzanne Wendelken

Spatial and temporal characteristics of V1 microstimulation during chronic implantation of a microelectrode array in a behaving macaque

OBJECTIVE It has been hypothesized that a vision prosthesis capable of evoking useful visual percepts can be based upon electrically stimulating the primary visual cortex (V1) of a blind human subject via penetrating microelectrode arrays. As a continuation of earlier work, we examined several spatial and temporal characteristics of V1 microstimulation. APPROACH An array of 100 penetrating microelectrodes was chronically implanted in V1 of a behaving macaque monkey. Microstimulation thresholds were measured using a two-alternative forced choice detection task. Relative locations of electrically-evoked percepts were measured using a memory saccade-to-target task. MAIN RESULTS The principal finding was that two years after implantation we were able to evoke behavioural responses to electric stimulation across the spatial extent of the array using groups of contiguous electrodes. Consistent responses to stimulation were evoked at an average threshold current per electrode of 204 ± 49 µA (mean ± std) for groups of four electrodes and 91 ± 25 µA for groups of nine electrodes. Saccades to electrically-evoked percepts using groups of nine electrodes showed that the animal could discriminate spatially distinct percepts with groups having an average separation of 1.6 ± 0.3 mm (mean ± std) in cortex and 1.0° ± 0.2° in visual space. Significance. These results demonstrate chronic perceptual functionality and provide evidence for the feasibility of a cortically-based vision prosthesis for the blind using penetrating microelectrodes.

Using multiple high-count electrode arrays in human median and ulnar nerves to restore sensorimotor function after previous transradial amputation of the hand.

Peripheral nerve interfaces that can record from and stimulate large numbers of different nerve fibers selectively and independently may help restore intuitive and effective motor and sensory function after hand amputation. To this end, and extending previous work in two subjects, two 100-electrode Utah Slanted Electrode Arrays (USEAs) were implanted for four weeks in the residual ulnar and median nerves of a 50-year-old male whose left, dominant hand had been amputated 21 years previously. Subsequent experiments involved 1) recording from USEAs for real-time control of a virtual prosthetic hand; 2) stimulation to evoke somatosensory percepts; and 3) closed-loop sensorimotor control. Overall, partial motor control and sensation were achieved using USEAs. 1) Isolated action potentials recorded from nerve motor fibers, although sparse at these distal implant sites, were activated during fictive movements of the phantom hand. Unlike in our previous two subjects, electromyographic (EMG) activity contributed to most online recordings and decodes, but was reduced in offline analyses using common average referencing. Online and offline Kalman-filter decodes of thresholded neural or EMG spikes independently controlled different digits of the virtual hand with one or two degrees of freedom. 2) Microstimulation through individual electrodes of the two USEAs evoked up to 106 different percepts, covering much of the phantom hand. The subject discriminated among five perceived stimulus locations, and between two somatosensory submodalities at a single location. 3) USEA-evoked percepts, mimicking contact with either a near or distal virtual target, were used to terminate movements of the virtual hand controlled with USEA recordings comprised wholly or mostly of EMG. These results further indicate that USEAs can help restore sensory and motor function after hand loss.

Recording and Decoding for Neural Prostheses

This paper reviews technologies and signal processing algorithms for decoding peripheral nerve and electrocorticogram signals to interpret human intent and control prosthetic arms. The review includes a discussion of human motor system physiology and physiological signals that can be used to decode motor intent, electrode technology for acquiring neural data, and signal processing methods including decoders based on Kalman filtering and least-squares regressors. Representative results from human experiments demonstrate the progress that has been made in neural decoding and its potential for developing neuroprosthetic arms that act and feel like natural arms.

Polynomial Kalman filter for myoelectric prosthetics using efficient kernel ridge regression

This paper presents a polynomial ridge regression algorithm with substantial improvements in computational efficiency compared with the polynomial kernel ridge regression and the standard polynomial regression. This regression algorithm was combined with a Kalman Filter (KF) to yield the Directly Weighted Polynomial Ridge Regression KF (DWPRR-KF). Experiments conducted offline from data collected from a human amputee demonstrated that compared with a linear KF, the DWPRR-KF significantly reduced median range-normalized Root Mean Square Error (RMSE) caused by movement on Degrees Of Freedom (DOFs) that the user intended to hold stationary during movement of other DOFs by 63% (from 0.061 to 0.023), while insignificantly increasing median error on DOFs the user intended to move (3%; from 0.139 to 0.144). Furthermore, the median overall error, from DOFs with or without intended movement, decreased by 27% (from 0.085 to 0.063) but this change was not found significant.

Individual hand movement detection and classification using peripheral nerve signals

This paper investigates whether the movement intent of an amputee can be detected and classified in real-time as the individual moved his/her phantom hand. We present a method to detect movement intent using neural signals from the peripheral nervous system (PNS). In addition, we classify eight types of individual hand movements using 300 ms signal segments beginning with our detected starting time. Classification is performed by applying linear discriminant analysis (LDA) on different kind of features. We compared the classification results using segments started with the detected starting time and the starting time of the command given to a subject as neural signals were recorded. The average accuracies were 73.5% in the former case and 59.4% in the latter.

Neural decoding systems using Markov Decision Processes

This paper presents a framework for modeling neural decoding using electromyogram (EMG) and electrocorticogram (ECoG) signals to interpret human intent and control prosthetic arms. Specifically, the method of this paper employs Markov Decision Processes (MDP) for neural decoding, parameterizing the policy using an artificial neural network. The system is trained using a modification of the Dataset Aggregation (DAgger) algorithm. The results presented here suggest that the approach of the paper performs better than the state-of-the-art.

Motor Control and Sensory Feedback Enhance Prosthesis Embodiment and Reduce Phantom Pain After Long-Term Hand Amputation

We quantified prosthesis embodiment and phantom pain reduction associated with motor control and sensory feedback from a prosthetic hand in one human with a long-term transradial amputation. Microelectrode arrays were implanted in the residual median and ulnar arm nerves and intramuscular electromyography recording leads were implanted in residual limb muscles to enable sensory feedback and motor control. Objective measures (proprioceptive drift) and subjective measures (survey answers) were used to assess prosthesis embodiment. For both measures, there was a significant level of embodiment of the physical prosthetic limb after open-loop motor control of the prosthesis (i.e., without sensory feedback), open-loop sensation from the prosthesis (i.e., without motor control), and closed-loop control of the prosthesis (i.e., motor control with sensory feedback). There was also a statistically significant reduction in reported phantom pain after experimental sessions that included open-loop nerve microstimulation, open-loop prosthesis motor control, or closed-loop prosthesis motor control. The closed-loop condition provided no additional significant improvements in phantom pain reduction or prosthesis embodiment relative to the open-loop sensory condition or the open-loop motor condition. This study represents the first long-term (14-month), systematic report of phantom pain reduction and prosthesis embodiment in a human amputee across a variety of prosthesis use cases.

Linear methods for reducing EMG contamination in peripheral nerve motor decodes

Signals recorded from the peripheral nervous system (PNS) with high channel count penetrating microelectrode arrays, such as the Utah Slanted Electrode Array (USEA), often have electromyographic (EMG) signals contaminating the neural signal. This common-mode signal source may prevent single neural units from successfully being detected, thus hindering motor decode algorithms. Reducing this EMG contamination may lead to more accurate motor decode performance. A virtual reference (VR), created by a weighted linear combination of signals from a subset of all available channels, can be used to reduce this EMG contamination. Four methods of determining individual channel weights and six different methods of selecting subsets of channels were investigated (24 different VR types in total). The methods of determining individual channel weights were equal weighting, regression-based weighting, and two different proximity-based weightings. The subsets of channels were selected by a radius-based criteria, such that a channel was included if it was within a particular radius of inclusion from the target channel. These six radii of inclusion were 1.5, 2.9, 3.2, 5, 8.4, and 12.8 electrode-distances; the 12.8 electrode radius includes all USEA electrodes. We found that application of a VR improves the detectability of neural events via increasing the SNR, but we found no statistically meaningful difference amongst the VR types we examined. The computational complexity of implementation varies with respect to the method of determining channel weights and the number of channels in a subset, but does not correlate with VR performance. Hence, we examined the computational costs of calculating and applying the VR and based on these criteria, we recommend an equal weighting method of assigning weights with a 3.2 electrode-distance radius of inclusion. Further, we found empirically that application of the recommended VR will require less than 1 ms for 33.3 ms of data from one USEA.