Larry Shupe - Academia.edu (original) (raw)

Papers by Larry Shupe

Supplementary figures in TIF format.

S1: Epoch based processing of ECoG power for classification of brain states. A) 10 seconds of raw... more S1: Epoch based processing of ECoG power for classification of brain states. A) 10 seconds of raw signal before VNS of one cortical site (on the left). The red vertical line indicates the end of stimulation (the black vertical lines are the stimulation artifacts). The two vertical dashed lines mark the 6-sec epoch used to generate the power spectrum density on the right. The blue dashed lines mark the limits of the frequency bands considered for the power analysis (δ = 1-4 Hz, θ = 4-8 Hz, α = 8-14, β = 14-35 Hz, γ = 35-55 Hz). B) Example of power values for each frequency band calculated from the epoch displayed in A. From left to right: absolute power, relative power and transformed relative power. C) Number of unclassified epochs as a function of different power threshold values used to discriminate between brain states. The threshold which returned the minimum number of unclassified epochs was chosen for the classification procedure. D) Example of power distributions in delta and...

S1: Comparison based on the number of epochs for each classified brain state and for each recordi... more S1: Comparison based on the number of epochs for each classified brain state and for each recording site between two different classifiers. The first classifier included the information about when the room light was turned ON and OFF (top row) while the second classifier did not include this criterion (second row). The scatter plots on the bottom compare for each channel (represented by a dot) the number of classified epochs between the two classifiers. Despite some minor differences the two classifiers returned similar number of epochs for all channels, as highlighted by the identity line. Red: active-wake; blue: resting-wake; green: REM, pink: NREM. S2: Averaged number of classified epochs over all channels (± SD) as function of different movement threshold (X axis). We chose 60% for both animals to classify movement vs. resting. At 60% the number of epochs of classified as AW and RW are the same for M2. For M1 the two lines crossed at 85%. The morphology of the classified VEPs wa...

We report the first EEG recordings of an expert in the 8 advanced meditations called jhanas, and ... more We report the first EEG recordings of an expert in the 8 advanced meditations called jhanas, and propose 5 hypotheses on how the jhanas differ in EEG power from the eyes-closed resting state at 7 different brain regions. We hypothesize simple changes in the brain regions responsible for each of the 5 principal experiential features of jhana states. These features are: (1) internal verbalizations fade, (2) external awareness dims, (3) the sense of personal boundaries is altered, (4) the experience of evaluations, goals, and "shoulds" diminishes, and (5) attention is highly focused on the object of meditation. The results strongly confirm reduced activity in the brain regions related to the first 3 hypotheses, with all 16 of the planned comparisons significant and in the predicted direction. With respect to Hypothesis 4, results are mixed, with all 4 predictions significantly confirmed in the alpha1 band, but all 4 disconfirmed in the theta band. Lastly, Hypothesis 5 was mos...

Running title: Vagal evoked potentials are modulated by brain state 7 8 Irene Rembado, David K. S... more Running title: Vagal evoked potentials are modulated by brain state 7 8 Irene Rembado, David K. Su, Ariel Levari, Larry E. Shupe, Steve Perlmutter, Eberhard Fetz, Stavros 9 Zanos 10 11 1Department of Physiology & Biophysics, University of Washington, 1705 NE Pacific St, Seattle, WA 12 98195, USA 13 2 Department of Neurological Surgery, University of Washington, 1959 NE Pacific St, Seattle, WA 14 98195, USA 15 3Institute of Bioelectronic Medicine, Feinstein Institutes for Medical Research, 350 Community Drive, 16 Manhasset NY 11030, USA 17 18 Corresponding authors: Irene Rembado: irene.rembado@gmail.com, Stavros Zanos: 19 szanos@northwell.edu 20

Cerebral Cortex, 2021

Vagus nerve stimulation (VNS) has been tested as therapy for several brain disorders and as a mea... more Vagus nerve stimulation (VNS) has been tested as therapy for several brain disorders and as a means to modulate cortical excitability and brain plasticity. Cortical effects of VNS, manifesting as vagal-evoked potentials (VEPs), are thought to arise from activation of ascending cholinergic and noradrenergic systems. However, it is unknown whether those effects are modulated by brain state at the time of stimulation. In 2 freely behaving macaque monkeys, we delivered short trains of 5 pulses to the left cervical vagus nerve at different frequencies (5-300 Hz) while recording local field potentials (LFPs) from sites in contralateral prefrontal, sensorimotor and parietal cortical areas. Brain states were inferred from spectral components of LFPs and the presence of overt movement: active awake, resting awake, REM sleep and NREM sleep. VNS elicited VEPs in all sampled cortical areas. VEPs comprised early (<70 ms), intermediate (70-250 ms) and late (>250 ms) components. The magnitud...

Frontiers in Neuroscience, 2021

Toward addressing many neuroprosthetic applications, the Neurochip3 (NC3) is a multichannel bidir... more Toward addressing many neuroprosthetic applications, the Neurochip3 (NC3) is a multichannel bidirectional brain-computer interface that operates autonomously and can support closed-loop activity-dependent stimulation. It consists of four circuit boards populated with off-the-shelf components and is sufficiently compact to be carried on the head of a non-human primate (NHP). NC3 has six main components: (1) an analog front-end with an Intan biophysical signal amplifier (16 differential or 32 single-ended channels) and a 3-axis accelerometer, (2) a digital control system comprised of a Cyclone V FPGA and Atmel SAM4 MCU, (3) a micro SD Card for 128 GB or more storage, (4) a 6-channel differential stimulator with ±60 V compliance, (5) a rechargeable battery pack supporting autonomous operation for up to 24 h and, (6) infrared transceiver and serial ports for communication. The NC3 and earlier versions have been successfully deployed in many closed-loop operations to induce synaptic plas...

We describe an integrate-and-fire (IF) spiking neural network that incorporates spike-timing depe... more We describe an integrate-and-fire (IF) spiking neural network that incorporates spike-timing dependent plasticity (STDP) and simulates the experimental outcomes of four different conditioning protocols that produce cortical plasticity. The original conditioning experiments were performed in freely moving non-human primates with an autonomous head-fixed bidirectional brain-computer interface. Three protocols involved closed-loop stimulation triggered from (a) spike activity of single cortical neurons, (b) EMG activity from forearm muscles, and (c) cycles of spontaneous cortical beta activity. A fourth protocol involved open-loop delivery of pairs of stimuli at neighboring cortical sites. The IF network that replicates the experimental results consists of 360 units with simulated membrane potentials produced by synaptic inputs and triggering a spike when reaching threshold. The 240 cortical units produce either excitatory or inhibitory post-synaptic potentials in their target units. I...

The Handbook of Brain Theory and Neural Networks, Oct 1, 1998

Neural plasticity, 2013

We report the first neural recording during ecstatic meditations called jhanas and test whether a... more We report the first neural recording during ecstatic meditations called jhanas and test whether a brain reward system plays a role in the joy reported. Jhanas are Altered States of Consciousness (ASC) that imply major brain changes based on subjective reports: (1) external awareness dims, (2) internal verbalizations fade, (3) the sense of personal boundaries is altered, (4) attention is highly focused on the object of meditation, and (5) joy increases to high levels. The fMRI and EEG results from an experienced meditator show changes in brain activity in 11 regions shown to be associated with the subjective reports, and these changes occur promptly after jhana is entered. In particular, the extreme joy is associated not only with activation of cortical processes but also with activation of the nucleus accumbens (NAc) in the dopamine/opioid reward system. We test three mechanisms by which the subject might stimulate his own reward system by external means and reject all three. Taken ...

The Journal of neuroscience : the official journal of the Society for Neuroscience, Jan 9, 2008

The pattern of remyelination after traumatic spinal cord injury remains elusive, with animal and ... more The pattern of remyelination after traumatic spinal cord injury remains elusive, with animal and human studies reporting partial to complete demyelination followed by incomplete remyelination. In the present study, we found that spared rubrospinal tract (RST) axons of passage traced with actively transported dextrans and examined caudally to the lesion 12 weeks after mouse spinal cord contusion injury were fully remyelinated. Spared axons exhibited a marginally reduced myelin thickness and significantly shorter internodes. CASPR (contactin-associated protein) and K(v)1.2 channels were used to identify internodes and paranodal protein distribution properties were used as an index of myelin integrity. This is the first time the CNS myelin internode length was measured in a mouse. To better understand the significance of shortened internodes and thinner myelin in spared axons, we modeled conduction properties using McIntyre's et al. model of myelinated axons. Mathematical modeling ...

1990 IJCNN International Joint Conference on Neural Networks, 1990

In monkeys performing a steptracking task, the discharge pattems of forearm motor units and conne... more In monkeys performing a steptracking task, the discharge pattems of forearm motor units and connected premotoneuronal cells in cortex and red nucleus (identified by post-spike facilitation of EMG) fall into characteristic classes: tonic, phasic-tonic, decrementing, etc. We used dynamic neural network models incorporating these discharge pattems to investigate networks that could transform a step input of target position to the observed discharge pattems of flexor and extensor motoneurons. These networks have interconnected hidden units with either excitatory or inhibitory connections to each other and to the motoneurons. The activity of many hidden units resembles discharge pattems that are observed in monkey recordings. The network solutions typically involve preferential connectivity within synergistic groups and often include reciprocal inhibition of antagonists. A network trained on a specific input step level does not necessarily produce a proportional output for other step sizes; however, the networks can be trained to generate motor responses proportional to a target step size. The role of the hidden units can also be investigated by selective lesions or stimulation.

Journal of Neuroscience, 2012

Remyelination following spinal cord injury (SCI) is thought to be incomplete; demyelination is re... more Remyelination following spinal cord injury (SCI) is thought to be incomplete; demyelination is reported to persist chronically and is proposed as a compelling therapeutic target. Yet most reports do not distinguish between the myelin status of intact axons and injurysevered axons whose proximal stumps persist but provide no meaningful function. We previously found full remyelination of spared, intact rubrospinal axons caudal to the lesion in chronic mouse SCI. However, the clinical concept of chronically demyelinated spared axons remains controversial. Since mouse models may have limitations in clinical translation, we asked whether the capacity for full remyelination is conserved in clinically relevant chronic rat SCI. We determined myelin status by examining paranodal protein distribution on anterogradely labeled, intact corticospinal and rubrospinal axons throughout the extent of the lesion. Demyelination was evident on proximal stumps of severed axons, but not on intact axons. For the first time, we demonstrate that a majority of intact axons exhibit remyelination (at least one abnormally short internode, Ͻ100 m). Remarkably, shortened internodes were significantly concentrated at the lesion epicenter and individual axons were thinned by 23% compared with their rostral and caudal zones. Mathematical modeling predicted a 25% decrease in conduction velocity at the lesion epicenter due to short internodes and axonal thinning. In conclusion, we do not find a large chronically demyelinated population to target with remyelination therapies. Interventions may be better focused on correcting structural or molecular abnormalities of regenerated myelin.

Science, 1994

al. present novel evidence which suggests that synaptic interactions between pairs of cortical ne... more al. present novel evidence which suggests that synaptic interactions between pairs of cortical neurons are directly related to the degree to which they fire together during directed limb movements (1) .The cover for the issue of 2

Neural Computation, 1994

Dynamic neural networks with recurrent connections were trained by backpropagation to generate th... more Dynamic neural networks with recurrent connections were trained by backpropagation to generate the differential or the leaky integral of a nonrepeating frequency-modulated sinusoidal signal. The trained networks performed these operations on arbitrary input waveforms. Reducing the network size by deleting ineffective hidden units and combining redundant units, and then retraining the network produced a minimal network that computed the same function and revealed the underlying computational algorithm. Networks could also be trained to compute simultaneously the differential and integral of the input on two outputs; the two operations were performed in distributed overlapping fashion, and the activations of the hidden units were dominated by the integral. Incorporating units with time constants into model networks generally enhanced their performance as integrators and interfered with their ability to differentiate.

Journal of Physiology-Paris, 2003

Dynamic recurrent neural networks composed of units with continuous activation functions provide ... more Dynamic recurrent neural networks composed of units with continuous activation functions provide a powerful tool for simulating a wide range of behaviors, since the requisite interconnections can be readily derived by gradient descent methods. However, it is not clear whether more realistic integrate-and-fire cells with comparable connection weights would perform the same functions. We therefore investigated methods to convert dynamic recurrent neural networks of continuous units into networks with integrateand-fire cells. The transforms were tested on two recurrent networks derived by backpropagation. The first simulates a short-term memory task with units that mimic neural activity observed in cortex of monkeys performing instructed delay tasks. The network utilizes recurrent connections to generate sustained activity that codes the remembered value of a transient cue. The second network simulates patterns of neural activity observed in monkeys performing a step-tracking task with flexion/extension wrist movements. This more complicated network provides a working model of the interactions between multiple spinal and supraspinal centers controlling motoneurons. Our conversion algorithm replaced each continuous unit with multiple integrate-and-fire cells that interact through delayed ''synaptic potentials''. Successful transformation depends on obtaining an appropriate fit between the activation function of the continuous units and the input-output relation of the spiking cells. This fit can be achieved by adapting the parameters of the synaptic potentials to replicate the input-output behavior of a standard sigmoidal activation function (shown for the short-term memory network). Alternatively, a customized activation function can be derived from the input-output relation of the spiking cells for a chosen set of parameters (demonstrated for the wrist flexion/extension network). In both cases the resulting networks of spiking cells exhibited activity that replicated the activity of corresponding continuous units. This confirms that the network solutions obtained through backpropagation apply to spiking networks and provides a useful method for deriving recurrent spiking networks performing a wide range of functions.

Journal of Computational Neuroscience, 2005

Dynamic recurrent neural networks were derived to simulate neuronal populations generating bidire... more Dynamic recurrent neural networks were derived to simulate neuronal populations generating bidirectional wrist movements in the monkey. The models incorporate anatomical connections of cortical and rubral neurons, muscle afferents, segmental interneurons and motoneurons; they also incorporate the response profiles of four populations of neurons observed in behaving monkeys. The networks were derived by gradient descent algorithms to generate the eight characteristic patterns of motor unit activations observed during alternating flexionextension wrist movements. The resulting model generated the appropriate input-output transforms and developed connection strengths resembling those in physiological pathways. We found that this network could be further trained to simulate additional tasks, such as experimentally observed reflex responses to limb perturbations that stretched or shortened the active muscles, and scaling of response amplitudes in proportion to inputs. In the final comprehensive network, motor units are driven by the combined activity of cortical, rubral, spinal and afferent units during step tracking and perturbations. The model displayed many emergent properties corresponding to physiological characteristics. The resulting neural network provides a working model of premotoneuronal circuitry and elucidates the neural mechanisms controlling motoneuron activity. It also predicts several features to be experimentally tested, for example the consequences of eliminating inhibitory connections in cortex and red nucleus. It also reveals that co-contraction can be achieved by simultaneous activation of the flexor and extensor circuits without invoking features specific to co-contraction.

IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2011

The Neurochip-2 is a second generation, battery-powered device for neural recording and stimulati... more The Neurochip-2 is a second generation, battery-powered device for neural recording and stimulating that is small enough to be carried in a chamber on a monkey's head. It has three recording channels, with user-adjustable gains, filters, and sampling rates, that can be optimized for recording single unit activity, local field potentials, electrocorticography, electromyography, arm acceleration, etc. Recorded data are stored on a removable, flash memory card. The Neurochip-2 also has three separate stimulation channels. Two "programmable-system-on-chips" (PSoCs) control the data acquisition and stimulus output. The PSoCs permit flexible real-time processing of the recorded data, such as digital filtering and time-amplitude window discrimination. The PSoCs can be programmed to deliver stimulation contingent on neural events or deliver preprogrammed stimuli. Access pins to the microcontroller are also available to connect external devices, such as accelerometers. The Neurochip-2 can record and stimulate autonomously for up to several days in freely behaving monkeys, enabling a wide range of novel neurophysiological and neuroengineering experiments.

Clinical Neurophysiology, 1999

Objective: We documented changes in spectral power of human electrocorticograms (ECoG) during per... more Objective: We documented changes in spectral power of human electrocorticograms (ECoG) during performance of sensorimotor tasks. Methods: In 6 human subjects, ECoGs were recorded simultaneously from 14 subdural cortical sites in forearm sensorimotor cortex. The subjects performed 3 visuomotor tasks: tracking a moving visual target with a joystick-controlled cursor, threading pieces of tubing, and pinching the ®ngers sequentially against the thumb. Control conditions consisted of passive resting and active extension of the wrist. For each site the spectral power of the ECoG during these behaviors was computed for 5 10 Hz ranges between 10 and 60 Hz. Results: All subjects showed power decreases in the range of 11±20 Hz and power increases in the 31±60 Hz range during performance of the visuomotor tasks, at sites in forearm sensorimotor cortex and adjacent areas. Simple wrist movements often produced little change in power. Three subjects showed episodes of explicit gamma oscillations during the visuomotor tasks. Different sites showed increases in gamma-range power for different tasks, indicating that the spatial distribution of the gamma activity is speci®c to the tasks. Cross-spectra showed that gamma activity could become synchronized between separate sites during particular tasks. Conclusions: Synchronized gamma-range activity in human sensorimotor cortex increases with performance of manipulative visuomotor tasks, supporting the hypothesis that coherent gamma oscillations may play a role in sensorimotor integration or attention.