Myoelectric interface for neurorehabilitation conditioning to reduce abnormal leg co-activation after stroke: a pilot study.

BACKGROUND: The ability to walk is an important factor in quality of life after stroke. Co-activation of hip adductors and knee extensors has been shown to correlate with gait impairment. We have shown previously that training with a myoelectric interface for neurorehabilitation (MINT) can reduce abnormal muscle co-activation in the arms of stroke survivors. METHODS: Here, we extend MINT conditioning to stroke survivors with leg impairment. The aim of this pilot study was to assess the safety and feasibility of using MINT to reduce abnormal co-activation between hip adductors and knee extensors and assess any effects on gait. Nine stroke survivors with moderate to severe gait impairment received 6 h of MINT conditioning over six sessions, either in the laboratory or at home. RESULTS: MINT participants completed a mean of 159 repetitions per session without any adverse events. Further, participants learned to isolate their muscles effectively, resulting in a mean reduction of co-activation of 70% compared to baseline. Moreover, gait speed increased by a mean of 0.15 m/s, more than the minimum clinically important difference. Knee flexion angle increased substantially, and hip circumduction decreased. CONCLUSION: MINT conditioning is safe, feasible at home, and enables reduction of co-activation in the leg. Further investigation of MINT's potential to improve leg movement and function after stroke is warranted. Abnormal co-activation of hip adductors and knee extensors may contribute to impaired gait after stroke. Trial registration This study was registered at ClinicalTrials.gov (NCT03401762, Registered 15 January 2018, https://clinicaltrials.gov/study/NCT03401762?tab=history&a=4 ).

Unsupervised Neural Manifold Alignment for Stable Decoding of Movement from Cortical Signals.

The stable decoding of movement parameters using neural activity is crucial for the success of brain-machine interfaces (BMIs). However, neural activity can be unstable over time, leading to changes in the parameters used for decoding movement, which can hinder accurate movement decoding. To tackle this issue, one approach is to transfer neural activity to a stable, low-dimensional manifold using dimensionality reduction techniques and align manifolds across sessions by maximizing correlations of the manifolds. However, the practical use of manifold stabilization techniques requires knowledge of the true subject intentions such as target direction or behavioral state. To overcome this limitation, an automatic unsupervised algorithm is proposed that determines movement target intention before manifold alignment in the presence of manifold rotation and scaling across sessions. This unsupervised algorithm is combined with a dimensionality reduction and alignment method to overcome decoder instabilities. The effectiveness of the BMI stabilizer method is represented by decoding the two-dimensional (2D) hand velocity of two rhesus macaque monkeys during a center-out-reaching movement task. The performance of the proposed method is evaluated using correlation coefficient and R-squared measures, demonstrating higher decoding performance compared to a state-of-the-art unsupervised BMI stabilizer. The results offer benefits for the automatic determination of movement intents in long-term BMI decoding. Overall, the proposed method offers a promising automatic solution for achieving stable and accurate movement decoding in BMI applications.

Myoelectric interface for neurorehabilitation conditioning to reduce abnormal leg co-activation after stroke: a pilot study.

Background: The ability to walk is an important factor in quality of life after stroke. Co-activation of hip adductors and knee extensors has been shown to correlate with gait impairment. We have shown previously that training with a myoelectric interface for neurorehabilitation (MINT) can reduce abnormal muscle co-activation in the arms of stroke survivors. Methods: Here, we extend MINT conditioning to stroke survivors with leg impairment. The aim of this pilot study was to assess the safety and feasibility of using MINT to reduce abnormal co-activation between hip adductors and knee extensors and assess any effects on gait. Nine stroke survivors with moderate to severe gait impairment received six hours of MINT conditioning over six sessions, either in the laboratory or at home. Results: MINT participants completed a mean of 159 repetitions per session without any adverse events. Further, participants learned to isolate their muscles effectively, resulting in a mean reduction of co-activation of 70% compared to baseline. Moreover, gait speed increased by a mean of 0.15 m/s, more than the minimum clinically important difference. Knee flexion angle increased substantially, and hip circumduction decreased. Conclusion: MINT conditioning is safe, feasible at home, and enables reduction of co-activation in the leg. Further investigation of MINT's potential to improve leg movement and function after stroke is warranted. Abnormal co-activation of hip adductors and knee extensors may contribute to impaired gait after stroke. Trial registration: This study was registered at ClinicalTrials.gov (NCT03401762, Registered 15 January 2018, https://clinicaltrials.gov/study/NCT03401762?tab=history&a=4).

Automated lever task with minimum antigravity movement for rats with cervical spinal cord injury.

BACKGROUND: Although there is currently no cure for paralysis due to spinal cord injury (SCI), the highest treatment priority is restoring arm and hand function for people with cervical SCI. Preclinical animal models provide an opportunity to test innovative treatments, but severe cervical injury models require significant time and effort to assess responses to novel interventions. Moreover, there is no behavioral task that can assess forelimb movement in rats with severe cervical SCI unable to perform antigravity movements. NEW METHOD: We developed a novel lever pressing task for rats with severe cervical SCI. We employed an automated adaptive algorithm to train animals using open-source software and commercially available hardware. We found that using the adaptive training required only 13.3 ± 2.5 training days to achieve behavioral proficiency. The lever press task could quantify immediate and long-term improvements in severely impaired forelimb function effectively. This behavior platform has potential to facilitate rehabilitative training and assess effects of therapeutic modalities following SCI. COMPARISON WITH EXISTING METHODS: There is no existing assessment aiming to quantify forelimb extension movement in rodents without function against gravity. We found that the new lever press task in the antigravity position could assess the severity of cervical SCI as well as the compensatory movement in the proximal forelimb less affected by the injury. CONCLUSIONS: This study demonstrates that the new behavioral task is capable of tracking the functional changes with various therapies in rats with severe forelimb impairments in a cost- and time-efficient manner.

Glassy Carbon Neural Interface for Chronic Epidural Stimulation in Rats with Cervical Spinal Cord Injury.

There is growing evidence on the efficacy of electrical stimulation delivered via spinal neural interfaces to improve functional recovery following spinal cord injury. For such interfaces, carbon-based neural arrays are fast becoming recognized as a compelling material and platform due to biocompatibility and long-term electrochemical stability. Here, we introduce the design, fabrication, and in vivo characterization of a novel cervical epidural implant with carbon-based surface electrodes. Through finite element analysis and mechanical load tests, we demonstrated that the array could safely withstand loads applied to it during implantation and natural movement of the subject with minimal stress levels. Furthermore, the long-term in vivo performance of this neural array consisting of glassy carbon surface electrodes was investigated through acute and chronic spinal motor evoked potential recordings and electrode impedance tests in rats. We demonstrated stable stimulation performance for at least four weeks in a rat model of spinal cord injury. Lastly, we found that impedance measurements on all carbon-based spinal arrays were generally stable over time up to four weeks after implantation, with a slight increase in impedance in subsequent weeks when tested in spinally injured rats. Taken together, this study demonstrated the potential for carbon-based electrodes as a spinal neural interface to accelerate both mechanistic research and functional restoration in animal models of spinal cord injury.

Brain-Computer-Spinal Interface Restores Upper Limb Function After Spinal Cord Injury.

Brain-computer interfaces (BCIs) are an emerging strategy for spinal cord injury (SCI) intervention that may be used to reanimate paralyzed limbs. This approach requires decoding movement intention from the brain to control movement-evoking stimulation. Common decoding methods use spike-sorting and require frequent calibration and high computational complexity. Furthermore, most applications of closed-loop stimulation act on peripheral nerves or muscles, resulting in rapid muscle fatigue. Here we show that a local field potential-based BCI can control spinal stimulation and improve forelimb function in rats with cervical SCI. We decoded forelimb movement via multi-channel local field potentials in the sensorimotor cortex using a canonical correlation analysis algorithm. We then used this decoded signal to trigger epidural spinal stimulation and restore forelimb movement. Finally, we implemented this closed-loop algorithm in a miniaturized onboard computing platform. This Brain-Computer-Spinal Interface (BCSI) utilized recording and stimulation approaches already used in separate human applications. Our goal was to demonstrate a potential neuroprosthetic intervention to improve function after upper extremity paralysis.

Adaptive Artifact Removal From Intracortical Channels for Accurate Decoding of a Force Signal in Freely Moving Rats.

Intracortical data recorded with multi-electrode arrays provide rich information about kinematic and kinetic states of movement in the brain-machine interface (BMI) systems. Direct estimation of kinetic information such as the force from cortical data has the same importance as kinematic information to make a functional BMI system. Various types of the information including single unit activity (SUA), multiunit activity (MUA) and local field potential (LFP) can be used as an input information to extract motor commands for control of the external devices in BMI. Here we combine LFP and MUA information to improve decoding accuracy of the force signal from the multi-channel intracortical data of freely moving rats. We suggest a weighted common average referencing (CAR) algorithm in order to valid interpretation of the force decoding from different data types. The proposed spatial filter adaptively identifies contribution of the common noise on the channels employing Kalman filter method. We evaluated the efficacy of the proposed artifact algorithm on both simulation and real data. In the simulation study, the average R 2 between the original and reconstructed signal of all channels after applying the proposed artifact removal method was computed for input SNRs in the range of -45 to 0 dB. Weighted CAR method can effectively reconstruct the original signal with average R 2 higher than 0.5 for input SNRs higher than -s10 dB in case of adding simulated outlier and motion artifacts. We also show that the proposed artifact removal algorithm 33% improves the accuracy of force decoding in terms of R 2 value compared to standard CAR filters.

Ultra-Capacitive Carbon Neural Probe Allows Simultaneous Long-Term Electrical Stimulations and High-Resolution Neurotransmitter Detection.

We present a new class of carbon-based neural probes that consist of homogeneous glassy carbon (GC) microelectrodes, interconnects and bump pads. These electrodes have purely capacitive behavior with exceptionally high charge storage capacity (CSC) and are capable of sustaining more than 3.5 billion cycles of bi-phasic pulses at charge density of 0.25 mC/cm2. These probes enable both high SNR (>16) electrical signal recording and remarkably high-resolution real-time neurotransmitter detection, on the same platform. Leveraging a new 2-step, double-sided pattern transfer method for GC structures, these probes allow extended long-term electrical stimulation with no electrode material corrosion. Cross-section characterization through FIB and SEM imaging demonstrate strong attachment enabled by hydroxyl and carbonyl covalent bonds between GC microstructures and top insulating and bottom substrate layers. Extensive in-vivo and in-vitro tests confirmed: (i) high SNR (>16) recordings, (ii) highest reported CSC for non-coated neural probe (61.4 ± 6.9 mC/cm2), (iii) high-resolution dopamine detection (10 nM level - one of the lowest reported so far), (iv) recording of both electrical and electrochemical signals, and (v) no failure after 3.5 billion cycles of pulses. Therefore, these probes offer a compelling multi-modal platform for long-term applications of neural probe technology in both experimental and clinical neuroscience.

Brain Control of an External Device by Extracting the Highest Force-Related Contents of Local Field Potentials in Freely Moving Rats.

A local field potential (LFP) signal is an alternative source to neural action potentials for decoding kinematic and kinetic information from the brain. Here, we demonstrate that the better extraction of force-related features from multichannel LFPs improves the accuracy of force decoding. We propose that applying canonical correlation analysis (CCA) filter on the envelopes of separate frequency bands (band-specific CCA) separates non-task related information from the LFPs. The decoding accuracy of the continuous force signal based on the proposed method were compared with three feature reduction methods: 1) band-specific principal component analysis (band-specific PCA) method that extract the components which leads to maximum variance from the envelopes of different frequency bands; 2) correlation coefficient-based (CC-based) feature reduction that selects the best features from the envelopes sorted based on the absolute correlation coefficient between each envelope and the target force signal; and 3) mutual information-based (MI-based) feature reduction that selects the best features from the envelopes sorted based on the mutual information between each envelope and output force signal. The band-specific CCA method outperformed band-specific PCA with 11% improvement, CC-based feature reduction with 16% improvement, and MI-based feature reduction with 18% improvement. In the online brain control experiments, the real-time decoded force signal from the 16-channel LFPs based on the proposed method was used to move a mechanical arm. Two rats performed 88 trials in seven sessions to control the mechanical arm based on the 16-channel LFPs.

Minimum Noise Estimate filter: a Novel Automated Artifacts Removal method for Field Potentials.

In this paper a novel automated and unsupervised method for removing artifacts from multichannel field potential signals is introduced which can be used in brain computer interface (BCI) applications. The method, which is called minimum noise estimate (MNE) filter is based on an iterative thresholding followed by Rayleigh quotient which tries to find an estimate of the noise and to minimize it over the original signal. MNE filter is capable to operate without any prior information about field potential signals. The performance of the proposed method is evaluated by its application on two different type of signals namely electrocorticogram (ECoG) and electroencephalogram (EEG) datasets through a decoding procedure. The results indicate that the proposed method outperforms well-known artifacts removal techniques such as common average referencing (CAR), Laplacian method, independent component analysis (ICA) and wavelet denoising approach.

Continuous Force Decoding from Local Field Potentials of the Primary Motor Cortex in Freely Moving Rats.

Local field potential (LFP) signals recorded by intracortical microelectrodes implanted in primary motor cortex can be used as a high informative input for decoding of motor functions. Recent studies show that different kinematic parameters such as position and velocity can be inferred from multiple LFP signals as precisely as spiking activities, however, continuous decoding of the force magnitude from the LFP signals in freely moving animals has remained an open problem. Here, we trained three rats to press a force sensor for getting a drop of water as a reward. A 16-channel micro-wire array was implanted in the primary motor cortex of each trained rat, and obtained LFP signals were used for decoding of the continuous values recorded by the force sensor. Average coefficient of correlation and the coefficient of determination between decoded and actual force signals were r = 0.66 and R2 = 0.42, respectively. We found that LFP signal on gamma frequency bands (30-120 Hz) had the most contribution in the trained decoding model. This study suggests the feasibility of using low number of LFP channels for the continuous force decoding in freely moving animals resembling BMI systems in real life applications.

Recognition of amyotrophic lateral sclerosis disease using factorial hidden Markov model.

Amyotrophic lateral sclerosis (ALS) is a common disease among neurological disorders that can change the pattern of gait in human. One of the effective methods for recognition and analysis of gait patterns in ALS patients is utilizing stride interval time series. With proper preprocessing for removing unwanted artifacts from the raw stride interval times and then extracting meaningful features from these data, the factorial hidden Markov model (FHMM) was used to distinguish ALS patients from healthy subjects. The results of classification accuracy evaluated using the leave-one-out (LOO) cross-validation algorithm showed that the FHMM method provides better recognition of ALS and healthy subjects compared to standard HMM. Moreover, comparing our method with a state-of-the art method named least square support vector machine (LS-SVM) showed the efficiency of the FHMM in distinguishing ALS subjects from healthy ones.

HMM for classification of Parkinson's disease based on the raw gait data.

The central nervous system (CNS) plays an important role in regulation of human gait. Parkinson's disease (PD) is a common neurodegenerative disease that may cause neurophysiologic change in the CNS and as a result change the gait cycle duration (stride interval). This article used the Hidden Markov Model (HMM) with Gaussian Mixtures to separate the patients with PD from healthy subjects. The results showed that the performance of the HMM classifier in classifying the gait data corresponding to 16 healthy and 15 PD subjects is comparable to the results obtained from the least squares support vector machine (LS-SVM) classifier. In this study, the leave-one-out cross-validation method was used to evaluate the performance of each classifier. The HMM method could effectively separate the gait data in terms of stride interval obtained from healthy subjects and PD patients with an accuracy rate of 90.3 % . All in all, the results showed that the proposed method can be used for distinguishing PD patients from healthy subjects based on the gait data classification.

Estimation of neural firing rate: the wavelet density estimation approach.

The computation of neural firing rates based on spike sequences has been introduced as a useful tool for extraction of an animal's behavior. Different estimating methods of such neural firing rates have been developed by neuroscientists, and among these methods, time histogram and kernel estimators have been used more than other approaches. In this paper, the problem in the estimation of firing rates using wavelet density estimators has been considered. The results of simulation study in estimation of underlying rates based on spike sequences sampled from two different variable firing rates show that the proposed wavelet density method provides better and more accurate estimation of firing rates with smooth results compared to two other classical approaches. Furthermore, the performance of a different family of wavelet density estimators in the estimation of the underlying firing rate of biological data have been compared with results of both time histogram and kernel estimators. All in all, the results show that the proposed method can be useful in the estimation of firing rate of neural spike trains.