Transfer Learning with CNN Models for Brain-Machine Interfaces to command lower-limb exoskeletons: A Solution for Limited Data.

This study evaluates the performance of two convolutional neural networks (CNNs) in a brain-machine interface (BMI) based on motor imagery (MI) by using a small dataset collected from five participants wearing a lower-limb exoskeleton. To address the issue of limited data availability, transfer learning was employed by training models on EEG signals from other subjects and subsequently fine-tuning them to specific users. A combination of common spatial patterns (CSP) and linear discriminant analysis (LDA) was used as a benchmark for comparison. The study's primary aim is to examine the potential of CNNs and transfer learning in the development of an automatic neural classification system for a BMI based on MI to command a lower-limb exoskeleton that can be used by individuals without specialized training.Clinical Relevance- BMI can be used in rehabilitation for patients with motor impairment by using mental simulation of movement to activate robotic exoskeletons. This can promote neural plasticity and aid in recovery.

Multisession, noninvasive closed-loop neuroprosthetic control of grasping by upper limb amputees.

Upper limb amputation results in a severe reduction in the quality of life of affected individuals due to their inability to easily perform activities of daily living. Brain-machine interfaces (BMIs) that translate grasping intent from the brain's neural activity into prosthetic control may increase the level of natural control currently available in myoelectric prostheses. Current BMI techniques demonstrate accurate arm position and single degree-of-freedom grasp control but are invasive and require daily recalibration. In this study we tested if transradial amputees (A1 and A2) could control grasp preshaping in a prosthetic device using a noninvasive electroencephalography (EEG)-based closed-loop BMI system. Participants attempted to grasp presented objects by controlling two grasping synergies, in 12 sessions performed over 5 weeks. Prior to closed-loop control, the first six sessions included a decoder calibration phase using action observation by the participants; thereafter, the decoder was fixed to examine neuroprosthetic performance in the absence of decoder recalibration. Ability of participants to control the prosthetic was measured by the success rate of grasping; ie, the percentage of trials within a session in which presented objects were successfully grasped. Participant A1 maintained a steady success rate (63±3%) across sessions (significantly above chance [41±5%] for 11 sessions). Participant A2, who was under the influence of pharmacological treatment for depression, hormone imbalance, pain management (for phantom pain as well as shoulder joint inflammation), and drug dependence, achieved a success rate of 32±2% across sessions (significantly above chance [27±5%] in only two sessions). EEG signal quality was stable across sessions, but the decoders created during the first six sessions showed variation, indicating EEG features relevant to decoding at a smaller timescale (100ms) may not be stable. Overall, our results show that (a) an EEG-based BMI for grasping is a feasible strategy for further investigation of prosthetic control by amputees, and (b) factors that may affect brain activity such as medication need further examination to improve accuracy and stability of BMI performance.

Brain-computer interface devices for patients with paralysis and amputation: a meeting report.

OBJECTIVE: The Food and Drug Administration's (FDA) Center for Devices and Radiological Health (CDRH) believes it is important to help stakeholders (e.g., manufacturers, health-care professionals, patients, patient advocates, academia, and other government agencies) navigate the regulatory landscape for medical devices. For innovative devices involving brain-computer interfaces, this is particularly important. APPROACH: Towards this goal, on 21 November, 2014, CDRH held an open public workshop on its White Oak, MD campus with the aim of fostering an open discussion on the scientific and clinical considerations associated with the development of brain-computer interface (BCI) devices, defined for the purposes of this workshop as neuroprostheses that interface with the central or peripheral nervous system to restore lost motor or sensory capabilities. MAIN RESULTS: This paper summarizes the presentations and discussions from that workshop. SIGNIFICANCE: CDRH plans to use this information to develop regulatory considerations that will promote innovation while maintaining appropriate patient protections. FDA plans to build on advances in regulatory science and input provided in this workshop to develop guidance that provides recommendations for premarket submissions for BCI devices. These proceedings will be a resource for the BCI community during the development of medical devices for consumers.

A model for altered neural network dynamics related to prehension movements in Parkinson disease.

In this paper, we present a neural network model of the interactions between cortex and the basal ganglia during prehensile movements. Computational neuroscience methods are used to explore the hypothesis that the altered kinematic patterns observed in Parkinson's disease patients performing prehensile movements is mainly due to an altered neuronal activity located in the networks of cholinergic (ACh) interneurons of the striatum. These striatal cells, under a strong influence of the dopaminergic system, significantly contribute to the neural processing within the striatum and in the cortico-basal ganglia loops. In order to test this hypothesis, a large-scale model of neural interactions in the basal ganglia has been integrated with previous models accounting for the cortical organization of goal directed reaching and grasping movements in normal and perturbed conditions. We carry out a discussion of the model hypothesis validation by providing a control engineering analysis and by comparing results of real experiments with our simulation results in conditions resembling these original experiments.

Abrupt, but not gradual visuomotor distortion facilitates adaptation in children with developmental coordination disorder.

A previous experiment investigating visuomotor adaptation in typically developing children and children with Developmental Coordination Disorder (DCD) suggested poor adaptation to an abruptly induced visuomotor perturbation. In the current study, using a similar center-out drawing task, but administering either an abrupt or a gradual perturbation, and twice as many adaptation trials, we show that typically developing children are well able to successfully update an existing internal model in response to a 60 degrees rotation of the visual feedback, independent of the perturbation condition. Children with DCD, however, updated their internal map more effectively during exposure to an abrupt visuomotor perturbation than to a gradual one. This may suggest that the adaptation process in children with DCD responds differently to small vs. large steps of visuomotor discrepancies. Given the known role of the cerebellum in providing an error signal necessary for updating the internal model in response to a gradual visuomotor distortion, the results of our study add to the growing body of evidence implicating compromised cerebellar function in DCD.

Effects of increased complexity of visuo-motor transformations on children's arm movements.

The effects of increasing complexity of visuo-motor transformations on movement were examined in 4-, 6-, and 8-year-old children and adults. Participants performed a 'center-out' drawing task under three increasingly complex conditions: (1) Normal transformation: The target, line path and hand position were fully visible, in the horizontal plane, throughout the movement. (2) Aligned transformation: The target and line path were displayed horizontally above the workspace, with vision of the arm/hand occluded. (3) Vertical transformation: The target and line paths were presented on a vertical computer monitor with vision of the arm/hand occluded. Results showed that with increasing age, movements became faster, straighter, and smoother. The 4- and 6-year-old children were more variable in their specification of movement direction than the 8-year-old children and the adults, and were also more affected by the complexity of the transformation. This suggested that besides the complexity of the visual transformation, the familiarity/experienced environment might also play a role in 'sharpening' the transformation maps represented in movement planning.

Adaptation of handwriting size under distorted visual feedback in patients with Parkinson's disease and elderly and young controls.

OBJECTIVE: The ability to use visual feedback to control handwriting size was compared in patients with Parkinson's disease (PD), elderly people, and young adults to better understand factors playing a part in parkinsonian micrographia. METHODS: The participants wrote sequences of eight cursive l loops with visual target sizes of 0.5 and 2 cm on a flat panel display digitiser which both recorded and displayed the pen movements. In the pre-exposure and postexposure conditions, the display digitiser showed the actual pen trace in real time and real size. In the distortion exposure conditions, the gain of the vertical dimension of the visual feedback was either reduced to 70% or enlarged to 140%. RESULTS: The young controls showed a gradual visuomotor adaptation that compensated for the visual feedback distortions during the exposure conditions. They also showed significant after effects during the postexposure conditions. The elderly controls marginally corrected for the size distortions and showed small after effects. The patients with PD, however, showed no trial by trial adaptations or after effects but instead, a progressive amplification of the distortion effect in each individual trial. CONCLUSION: The young controls used visual feedback to update their visuomotor map. The elderly controls seemed to make little use of visual feedback. The patients with Parkinson's disease rely on the visual feedback of previous or of ongoing strokes to programme subsequent strokes. This recursive feedback may play a part in the progressive reductions in handwriting size found in parkinsonian micrographia.

Parkinson's disease and the control of size and speed in handwriting.

This experiment investigated whether Parkinson's disease (PD) patients experience problems in producing stroke size, stroke duration or both, in a handwriting task. Thirteen PD patients and 15 elderly controls wrote four patterns of varying complexity on a digitizer tablet. The participants were instructed to execute the writing movements: at a normal size and speed; as fast as possible; two times larger than normal; and two times larger and as fast as possible. PD patients had no difficulty increasing speed while maintaining size and had no difficulty increasing size while maintaining speed. However, they showed significantly smaller size increases in the two times larger condition as compared to the elderly controls. The conditions were also simulated by a neural network model of normal and PD movement control that produced a stroke pattern that approximated the experimental data. For the instructions used, the results suggest that when patients scale speed, they have no difficulty controlling force amplitude, but when they scale stroke size, they have a problem controlling force amplitude. Thus, PD patients may have reduced capability to maintain a given force level for the stroke time periods tested with the instructions.

A predictive reinforcement model of dopamine neurons for learning approach behavior.

A neural network model of how dopamine and prefrontal cortex activity guides short- and long-term information processing within the cortico-striatal circuits during reward-related learning of approach behavior is proposed. The model predicts two types of reward-related neuronal responses generated during learning: (1) cell activity signaling errors in the prediction of the expected time of reward delivery and (2) neural activations coding for errors in the prediction of the amount and type of reward or stimulus expectancies. The former type of signal is consistent with the responses of dopaminergic neurons, while the latter signal is consistent with reward expectancy responses reported in the prefrontal cortex. It is shown that a neural network architecture that satisfies the design principles of the adaptive resonance theory of Carpenter and Grossberg (1987) can account for the dopamine responses to novelty, generalization, and discrimination of appetitive and aversive stimuli. These hypotheses are scrutinized via simulations of the model in relation to the delivery of free food outside a task, the timed contingent delivery of appetitive and aversive stimuli, and an asymmetric, instructed delay response task.

Visuo-motor adaptation in smokeless tobacco users.

Ten smokeless tobacco (ST) users and 11 non-smokers participated in a visuo-motor adaptation experiment in which the visual feedback of point-to-point horizontal arm movements, displayed in real-time on a computer screen, was rotated by 45 degrees counterclockwise for some trials. Visuo-motor performance between smokers and non-smokers was compared on three occasions, once after at least 8 h of tobacco abstinence (Session 1), a second time following ST intake (Session 2), and a third time 45 min after the original ST intake (Session 3). Non-smokers were tested at the same relative times as the smokers in the absence of any tobacco. Both groups performed the three conditions during each session: baseline (normal visual feedback), rotated visual feedback (45 degrees visual feedback rotation), and post-adaptation (normal visual feedback immediately following feedback rotation). Compared with non-smokers, ST users had significantly larger normalized jerk scores (a measure of movement smoothness) after ST intake during the adaptation and post-adaptation conditions in Sessions 2 and 3, but not during the baseline conditions, implying a differential effect of ST use specific to rotated visual feedback. Movement duration was also longer for smokers than for non-smokers after ST intake, but only in the post-adaptation condition. Overall the results suggest that ST use, and hence nicotine, has a detrimental effect on visuo-motor performance, particularly on movement smoothness.

Elderly subjects are impaired in spatial coordination in fine motor control.

Elderly and young control subjects performed back-and-forth handwriting movements in various orientations, therefore varying the coordination demands. Elderly subjects showed higher normalized jerk and straightness scores than the young subjects. However, jerk scores were independent of the coordination demands in either group. In contrast, the straightness scores were highly dependent on stroke orientation for the elderly, but they remained constant across orientations for the young controls. Moreover, group differences in stroke size and stroke duration were not significant, and orientation effects were unrelated. It is suggested that the orientation-dependent straightness scores in the elderly may result from unequal timing or improper scaling of muscle forces. These data suggest that aging deteriorates the spatial coordination of finger and wrist movements, but not accelerative force control.

Neural dynamics of short and medium-term motor control effects of levodopa therapy in Parkinson's disease.

A neural network model of movement control in normal and Parkinson's disease (PD) conditions is proposed to simulate the time-varying dose-response relationship underlying the effects of levodopa on movement amplitude and movement duration in PD patients. Short and long-term dynamics of cell activations and neurotransmitter mechanisms underlying the differential expression of neuropeptide messenger RNA within the basal ganglia striatum are modeled to provide a mechanistic account for the effects of levodopa medication on motor performance (e.g. the pharmacodynamics). Experimental and neural network simulation data suggest that levodopa therapy in Parkinson's disease has differential effects on cell activities, striatal neuropeptides, and motor behavior. In particular, it is shown how dopamine depletion in the striatum may modulate differentially the level of substance P and enkephalin messenger RNA in the direct and indirect basal ganglia pathways. This dissociation in the magnitude and timing of peptide expression causes an imbalance in the opponently organized basal ganglia pathways which results in Parkinsonian motor deficits. The model is validated with experimental data obtained from handwriting movements performed by PD subjects before and after medication intake. The results suggest that fine motor control analysis and network modeling of the effects of dopamine in motor control are useful tools in drug development and in the optimization of pharmacological therapy in PD patients.

Parkinsonism reduces coordination of fingers, wrist, and arm in fine motor control.

This experiment investigates movement coordination in Parkinson's disease (PD) subjects. Seventeen PD patients and 12 elderly control subjects performed several handwriting-like tasks on a digitizing writing tablet resting on top of a table in front of the subject. The writing patterns, in increasing order of coordination complexity, were repetitive back-and-forth movements in various orientations, circles and loops in clockwise and counterclockwise directions, and a complex writing pattern. The patterns were analyzed in terms of jerk normalized for duration and size per stroke. In the PD subjects, back-and-forth strokes, involving coordination of fingers and wrist, showed larger normalized jerk than strokes performed using either the wrist or the fingers alone. In the PD patients, wrist flexion (plus radial deviation) showed greater normalized jerk in comparison to wrist extension (plus ulnar deviation). The elderly control subjects showed no such effects as a function of coordination complexity. For both PD and elderly control subjects, looping patterns consisting of circles with a left-to-right forearm movement, did not show a systematic increase of normalized jerk. The same handwriting patterns were then simulated using a biologically inspired neural network model of the basal ganglia thalamocortical relations for a control and a mild PD subject. The network simulation was consistent with the observed experimental results, providing additional support that a reduced capability to coordinate wrist and finger movements may be caused by suboptimal functioning of the basal ganglia in PD. The results suggest that in PD patients fine motor control problems may be caused by a reduced capability to coordinate the fingers and wrist and by reduced control of wrist flexion.

Adaptation to gradual as compared with sudden visuo-motor distortions.

If visual feedback is discordant with movement direction, the visuo-motor mapping is disrupted, but can be updated with practice. In this experiment subjects practiced discrete arm movements under conditions of visual feedback rotation. One group was exposed to 10 degree-step increments of visual feedback rotation up to a total of 90 degrees, a second group to a 90 degree visual feedback rotation throughout the experiment. After the first group reached the 90 degree visual feedback rotation, its subjects performed faster, with less spatial error, and showed larger aftereffects than the subjects who practiced constantly under the 90 degree visual feedback rotation condition. Results suggest that gradually increasing feedback distortion allows more complete adaptation than a large, sudden distortion onset.

A neural model of cerebellar learning for arm movement control: cortico-spino-cerebellar dynamics.

A neural network model of opponent cerebellar learning for arm movement control is proposed. The model illustrates how a central pattern generator in cortex and basal ganglia, a neuromuscular force controller in spinal cord, and an adaptive cerebellum cooperate to reduce motor variability during multijoint arm movements using mono- and bi-articular muscles. Cerebellar learning modifies velocity commands to produce phasic antagonist bursts at interpositus nucleus cells whose feed-forward action overcomes inherent limitations of spinal feedback control of tracking. Excitation of alpha motoneuron pools, combined with inhibition of their Renshaw cells by the cerebellum, facilitate movement initiation and optimal execution. Transcerebellar pathways are opened by learning through long-term depression (LTD) of parallel fiber-Purkinje cell synapses in response to conjunctive stimulation of parallel fibers and climbing fiber discharges that signal muscle stretch errors. The cerebellar circuitry also learns to control opponent muscles pairs, allowing cocontraction and reciprocal inhibition of muscles. Learning is stable, exhibits load compensation properties, and generalizes better across movement speeds if motoneuron pools obey the size principle. The intermittency of climbing fiber discharges maintains stable learning. Long-term potentiation (LTP) in response to uncorrelated parallel fiber signals enables previously weakened synapses to recover. Loss of climbing fibers, in the presence of LTP, can erode normal opponent signal processing. Simulated lesions of the cerebellar network reproduce symptoms of cerebellar disease, including sluggish movement onsets, poor execution of multijoint plans, and abnormally prolonged endpoint oscillations.

Effects of Parkinsonism on motor control.

Parkinson's disease, which pathology results predominantly from nigros triatal pathway damage, has been associated with motor dysfunction due to basal ganglia impairment. It is argued that the variability seen within and between individual patients through the course of this neurological disorder may result from abnormal non-uniform neurotransmitter levels as well as functional segregation of neural populations in the basal ganglia. We review evidence that the wide spectrum of motor impairments observed in Parkinsonism may be due to a reduced capability of neurochemical modulation of pallido-thalamocortical activities that impairs movement implementation and execution.

Micrographia in Parkinson's disease.

A computational neural model of movement production in normal and Parkinson's disease (PD) is used to provide a neural account for the source of micrographia in PD handwriting. It is hypothesized that smaller than normal pallido-thalamic signals, due to dopamine depletion, are responsible for the observed overall smallness, slowness and variability in PD handwriting. Experimental data from PD patients that show micrographia support this hypothesis and imply the functional segregation of basal ganglia neural populations.

A neural model of basal ganglia-thalamocortical relations in normal and parkinsonian movement.

Anatomical, neurophysiological, and neurochemical evidence supports the notion of parallel basal ganglia-thalamocortical motor systems. We developed a neural network model for the functioning of these systems during normal and parkinsonian movement. Parkinson's disease (PD), which results predominantly from nigrostriatal pathway damage, is used as a window to examine basal ganglia function. Simulations of dopamine depletion produce motor impairments consistent with motor deficits observed in PD that suggest the basal ganglia play a role in motor initiation and execution, and sequencing of motor programs. Stereotaxic lesions in the model's globus pallidus and subthalamic nucleus suggest that these lesions, although reducing some PD symptoms, may constrain the repertoire of available movements. It is proposed that paradoxical observations of basal ganglia responses reported in the literature may result from regional functional neuronal specialization, and the non-uniform distributions of neurochemicals in the basal ganglia. It is hypothesized that dopamine depletion produces smaller-than-normal pallidothalamic gating signals that prevent rescalability of these signals to control variable movement speed, and that in PD can produce smaller-than-normal movement amplitudes.