https://deepblue.lib.umich.edu/bitstream/handle/2027.42/90300/j.ejpain.2004.11.001.pdf?sequence=1

Human brain mechanisms of pain perception and regulation in health and disease.

This is a proposal of the Movement Disorder Society for a clinical classification of tremors. The classification is based on the distinction between rest, postural, simple kinetic, and intention tremor (tremor during target-directed movements). Additional data from a medical history and the results of a neurologic examination can be combined into one of the following clinical syndromes defined in this statement: enhanced physiologic tremor, classical essential tremor (ET), primary orthostatic tremor, task- and position-specific tremors, dystonic tremor, tremor in Parkinson's disease (PD), cerebellar tremor, Holmes' tremor, palatal tremor, drug-induced and toxic tremor, tremor in peripheral neuropathies, or psychogenic tremor. Conditions such as asterixis, epilepsia partialis continua, clonus, and rhythmic myoclonus can be misinterpreted as tremor. The features distinguishing these conditions from tremor are described. Controversial issues are outlined in a comment section for each item and thus reflect the open questions that at present cannot be answered on a scientific basis. We hope that this statement provides a basis for better communication among clinicians working in the field and stimulates tremor research.

Consensus statement of the Movement Disorder Society on tremor

Objective Brain-computer interfaces (BCI) enable direct communication with a computer, using neural activity as the control signal. This neural signal is generally chosen from a variety of well-studied electroencephalogram (EEG) signals. For a given BCI paradigm, feature extractors and classifiers are tailored to the distinct characteristics of its expected EEG control signal, limiting its application to that specific signal. Convolutional neural networks (CNNs), which have been used in computer vision and speech recognition to perform automatic feature extraction and classification, have successfully been applied to EEG-based BCIs; however, they have mainly been applied to single BCI paradigms and thus it remains unclear how these architectures generalize to other paradigms. Here, we ask if we can design a single CNN architecture to accurately classify EEG signals from different BCI paradigms, while simultaneously being as compact as possible. Approach In this work we introduce EEGNet, a compact convolutional neural network for EEG-based BCIs. We introduce the use of depthwise and separable convolutions to construct an EEG-specific model which encapsulates well-known EEG feature extraction concepts for BCI. We compare EEGNet, both for within-subject and cross-subject classification, to current state-of-the-art approaches across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR). Main results We show that EEGNet generalizes across paradigms better than, and achieves comparably high performance to, the reference algorithms when only limited training data is available across all tested paradigms. In addition, we demonstrate three different approaches to visualize the contents of a trained EEGNet model to enable interpretation of the learned features. Significance Our results suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks. Our models can be found at: https://github.com/vlawhern/arl-eegmodels.

https://iopscience.iop.org/article/10.1088/1741-2552/aace8c/pdf

EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces

Brain computer interfaces (BCI) enable direct communication with a computer, using neural activity as the control signal. This neural signal is generally chosen from a variety of well-studied electroencephalogram (EEG) signals. For a given BCI paradigm, feature extractors and classifiers are tailored to the distinct characteristics of its expected EEG control signal, limiting its application to that specific signal. Convolutional Neural Networks (CNNs), which have been used in computer vision and speech recognition, have successfully been applied to EEG-based BCIs; however, they have mainly been applied to single BCI paradigms and thus it remains unclear how these architectures generalize to other paradigms. Here, we ask if we can design a single CNN architecture to accurately classify EEG signals from different BCI paradigms, while simultaneously being as compact as possible. In this work we introduce EEGNet, a compact convolutional network for EEG-based BCIs. We introduce the use of depthwise and separable convolutions to construct an EEG-specific model which encapsulates well-known EEG feature extraction concepts for BCI. We compare EEGNet to current state-of-the-art approaches across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR). We show that EEGNet generalizes across paradigms better than the reference algorithms when only limited training data is available. We demonstrate three different approaches to visualize the contents of a trained EEGNet model to enable interpretation of the learned features. Our results suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks, suggesting that the observed performances were not due to artifact or noise sources in the data.

EEGNet: A Compact Convolutional Network for EEG-based Brain-Computer Interfaces

What is the Bereitschaftspotential

Movement-related cortical potentials (MRPs) were recorded from scalp electrodes in 8 normal volunteers and from chronically implanted subdural electrodes in 7 patients who were being evaluated for surgical treatment of epilepsy. From subdural electrodes, a clearly defined, extremely localized slow negative potential preceding the voluntary movement of the middle finger (Bereitschaftspotential, BP) was recorded in the contralateral and ipsilateral hand sensorimotor areas. The negative slope (NS') began approximately 250 to 400 ms before EMG onset and was recorded exclusively from the contralateral hand sensorimotor area. Both BP and NS' were maximum in the hand motor area. Although a negative slope was recorded also from the supplementary motor area, whether that particular slope corresponded to BP or NS', or both, could not be determined. Three kinds of progressively steeper negative potentials starting around the onset of the EMG were identified: (1) the 'hand motor potentials' which were seen in the contralateral hand motor area and started immediately before EMG onset and peaked 130 +/- 32 ms after EMG onset; (2) the 'hand somatosensory potentials' seen in the contralateral hand somatosensory area which started simultaneously or immediately after the EMG onset; and (3) the 'vicinity potentials' seen in the immediate surroundings of the contralateral hand area and which started after the EMG onset. The 'hand motor potentials' had the highest amplitude. From these findings, we concluded that bilateral hand sensorimotor areas and the supplementary motor area participate in the 'preparation' of movements, but that mainly the contralateral cortex generates the discharges necessary to produce the actual movement.

Recording of movement-related potentials from scalp and cortex in man.

Fifty-seven consecutive patients with myoclonus from various causes were studied by electrophysiological techniques Giant somatosensory evoked potentials (SEPs) were observed almost exclusively in patients with progressive myoclonic epilepsy (PME) and diseases with similar clinical features that included lipidosis, neuronal ceroid lipofuscinosis and posthypoxic myoclonus On the basis of combinations of the giant SEP and the myoclonus-related cortical spike demonstrated by jerk-locked averaging, myoclonus in these patients was classified into four types In patients with 'cortical reflex' myoclonus (type I) who showed both the giant SEP and the myoclonus-related cortical spike, these two cortical activities were similar in terms of wave form, scalp topography, time relationship to either the long latency (C) reflex or myoclonus, the following cortical excitability, the effect of antimyoclonus drugs and alterations during slow wave sleep It is therefore postulated that the giant SEP is generated, at least in part, by common physiological mechanisms to the myoclonus-related cortical spike, or that the latter may comprise a constituent of the former In most patients with PME or allied diseases, both afferent and efferent components of the SEP are enhanced, but in some patients, only one of the two components seems to be predominantly enhanced

Pathogenesis of giant somatosensory evoked potentials in progressive myoclonic epilepsy.

Two patients with action tremor that was thought to originate in the cerebral cortex showed fine shivering-like finger twitching provoked mainly by action and posture. Surface EMG showed relatively rhythmic discharge at a rate of about 9 Hz, which resembled essential tremor. However, electrophysiologic studies revealed giant somatosensory evoked potentials (SEPs) with enhanced long-loop reflex and premovement cortical spike by the jerk-locked averaging method. Treatment with beta-blocker showed no effect, but anticonvulsants such as clonazepam, valproate, and primidone were effective to suppress the tremor and the amplitude of SEPs. We call this involuntary movement "cortical tremor," which is in fact a variant of cortical reflex myoclonus.

Cortical tremor: a variant of cortical reflex myoclonus.

Human auditory and somatosensory event-related potentials: effects of response condition and age.

Fifty-five consecutive cases of myoclonus owing to various etiologies were studied by conventional EEG-EMG polygraphic recordings and/or jerk-locked or back averaging. The technique of back-averaging was shown to be useful not only for detecting EEG correlates of myoclonus that are not recognizable on the routine polygraph but also for investigating the temporal and topographic relationship between the EEG activities and myoclonus. Thirteen of 17 cases of PME and related disorders, in whom back-averaging and SEP were studied, were shown to have both a myoclonus-related cortical spike over the contralateral central area, preceding the myoclonus of an upper extremity by 6 to 22 msec, and a giant SEP accompanied by an enhanced C reflex. In these cases of "cortical reflex myoclonus," the myoclonus-related spike was similar to the P25-N33 components of the giant SEP in its wave form, scalp topography, temporal relationship to myoclonus or to C reflex, succeeding cortical excitability, and drug effect. All of this suggests participation of common physiological mechanisms in those two activities. In two cases of PME, in which myoclonus involved bilateral proximal muscles synchronously, the myoclonus-related spike was maximal near the vertex, and there was no giant SEP. The significance of this subgroup remains undetermined. In six cases of the PME group, back-averaging was inapplicable because of rare occurrence of myoclonus, but they showed a typical giant SEP accompanied by an enhanced C reflex. In CJD, back-averaging demonstrated a sharp wave or PSD over the contralateral hemisphere, preceding the myoclonus by 50 to 85 msec. This form of myoclonus seems to be subcortical in origin. In essential myoclonus and oculopalatal-somatic myoclonus, there was neither myoclonus-related cortical spike nor giant SEP. Electrical stimulation of the peripheral nerve at variable intervals after the myoclonus onset (jerk-locked-SEP paradigm) was shown to be useful for investigating the influence of myoclonus on cortical excitability.

Ryuji Neshige

Papers

Recording of movement-related potentials from scalp and cortex in man.

Pathogenesis of giant somatosensory evoked potentials in progressive myoclonic epilepsy.

Cortical tremor: a variant of cortical reflex myoclonus.

Human auditory and somatosensory event-related potentials: effects of response condition and age.

Electroencephalographic correlates of myoclonus.