Search

This paper summarizes ongoing efforts in our laboratory to develop next-generation implantable neuroprobe systems that have closely integrated chemical and electrical neural interfaces. The MEMS-based neuroprobe consists of a silicon or polymer substrate having multiple metal sites for electrical recording and/or stimulation and one or more microchannels for fluid delivery. A chronoamperometric chemical sensor is achieved by selectively coating metal sites with an ion selective membrane. This multifunctional neuroprobe system provides a novel technology for establishing chronic and controllable chemical and electrical interfaces to targeted areas of the brain and spinal cord.

The capability of sensing catecholamine fluctuations in the brain is useful to determine how neurostimulation - both chemical and electrical - can affect catecholamine release and uptake. Here we report preliminary results on two multichannel neural probes capable of sensing dopamine through chronoamperometry methods. With a 16-channel microelectrode array and five-channel Michigan puffer probe with a microfluidic channel, we found that a 5-cycle dip coating process of Nafion, an ion-selective polymer, optimized sensitivity to dopamine and selectivity over ascorbic acid. In addition, through fluorescent microscopy, we have shown the suitability of Nafion as a "biocompatible" material for chronically implanted microelectrode probes. These findings support the ongoing work to develop chronic implantable chemical sensors.

A chief concern in the pursuit of controlling a neuroprosthetic device using direct brain signals is the question of how many bits of information are achievable through a direct brain-machine interface (BMI) via implantable microelectrode devices. This experiment begins to address this issue with implementation of a simple, software based decoding algorithm that allows the brain to adapt to the rules imposed upon it, To test this algorithm, two chronic 16-channel Michigan silicon microelectrode arrays were implanted into the primary motor cortex of two rats to record simultaneous unit spike activity. The animals were trained to perform an auditory detection task by modulating the recorded cortical spike activity in a prescribed manner. Both non-adaptive and adaptive neural decoding algorithms were evaluated. With the implementation of a non-adaptive decoding algorithm, the rats' behavioral (cortical) responses plateaued at approximately 75% correct; however, with the implementation of an adaptive algorithm, the rats' behavioral responses relatively quickly increased to 91% correct. The neural recordings suggest that the brain is able to modulate detailed cortical responses in accordance with the prescribed operant conditioning rules.

Surgical techniques are a critical contributor to the level of success achieved with chronically implanted cortical neuroprosthetic devices. Many different factors contribute to the amount of irritation the tissue is exposed to from the implanted device. Factors include mechanical irritations, infectious pathogens, dural regrowth, etc. In this paper we describe a novel application of the hydrogel polymer, ALGEL/spl reg/ (Neural Intervention Technologies, Ann Arbor, MI), in conjunction with an implanted Michigan probe (CNCT, University of Michigan). This polymer contains many inherent properties that are beneficial to this type of procedure. Properties include: 1) ease of application, 2) biocompatibility, 3) exemplary mechanical properties, and 4) translucent clarity. We believe that ALGEL has been a large contributor to the high level of success achieved with our chronic electrode implantations.

The formal characterization of the performance of microfluidic delivery devices is crucial for reliable in vivo application. A comprehensive laboratory technique was developed and used to optimize, calibrate and validate microfabricated fluid delivery devices. In vivo experiments were carried out to verify the accuracy and reliability of the pressure driven devices. Acute guinea pig experiments were conducted to measure the response to /spl alpha/-amino-3-hydroxy-5-methyl-4-isoxalone propionic acid, an excitatory neurotransmitter, at multiple locations in the inferior colliculus. A nondimensional parameter, Q/spl tilde/, was successfully used to classify devices in terms of geometry alone (i.e., independent of fluid properties). Functional devices exhibited long-term linearity and reliability in delivering single phase, Newtonian fluids, in discrete volumes with a resolution of 500 picoliters at less than 0.45 lbf/in/sup 2/ (30 mbar) pressure drop. Results for non-Newtonian fluids are not presented here. The acute results showed a proportional increase in the firing rate for delivered volumes of 2 nL up to 10 nL (at rates of between 0.1 and 1 nL/s). Flow characteristics are maintained during acute experiments and post-implant. A control experiment conducted with Ringer solution produced negligible effects, suggesting the results to be truly pharmacological. The experimental techniques employed have proven to be reliable and will be used for future calibration and testing of next generation chronic microfluidic delivery devices.

Construction of a direct brain-machine interface (BMI) for neuroprosthetic purposes is at the forefront of many current neural engineering thrusts. Due to recent breakthroughs in device technology and implantation techniques, a basic framework is now sufficiently developed to allow design of systems level interface strategies producing robust, scalable BMIs that adapt quickly to optimize information transfer at the interface. It has been postulated that knowledge of the underlying neural coding is mandatory for further BMI development. In this preliminary report we use an adaptive algorithm requiring limited knowledge of the underlying neural coding to allow naive rats implanted with Michigan silicon microelectrode arrays in motor cortex to perform a tone discrimination task via differential modulation of the recorded signals. One subject was able to perform the task consistently above chance, despite minor daily fluctuations in recording populations and signal quality. The brain rapidly changed response strategies to facilitate performance of the task, and the algorithm subsequently adapted to accommodate improved BMI operation.

The desire to perform chronic recordings from many channels in cortex is high among both neuroscientists and neural prosthesis researchers. The Michigan probe with integrated silicon ribbon cable has been used successfully for obtaining chronic cortical recordings in rodents. As these silicon devices are applied in larger animals and potentially in humans, however, a more mechanically robust, scalable system is required. Here we present a device that takes advantage of several existing, well-characterized technologies and methods and combines them to form a hybrid cortical assembly. The components of the assembly include thin-film multichannel silicon probes and polyimide cables that are electrically and mechanically coupled using thermocompression ball bonding. The resulting probe/cable assembly is low profile (

BioMEMS devices can be designed to provide viable neural interfaces for long-term, high-density, two-way communication with selected areas of cerebral cortex. Prototype thin-film polymer implantable microelectrode arrays were developed to extend the microelectrode design space in several ways, including enhanced flexibility, engineered surfaces and coatings, and new types of microchannels. Prototype MEMS silicon microdevices were developed as microsurgical tools for reliably inserting the flexible polymer electrodes into the cerebral cortex. Hybrid polymer microdevices were also developed for neural recording and stimulation combined with micro-drug delivery.

A combination of electrophysiological mapping, behavioral analysis and cortical micro-stimulation was used to explore the interrelation between the auditory cortex and behavior in the adult rat. Auditory discriminations were evaluated in eight rats trained to discriminate the presence or absence of a 75 dB pure tone stimulus. A probe trial technique was used to obtain intensity generalization gradients that described response probabilities to mid-level tones between 0 and 75 dB. The same rats were then chronically implanted in the auditory cortex with a 16 or 32 channel tungsten microwire electrode array. Implanted animals were then trained to discriminate the presence of single electrode micro-stimulation of magnitude 90 microA (22.5 nC/phase). Intensity generalization gradients were created to obtain the response probabilities to mid-level current magnitudes ranging from 0 to 90 microA on 36 different electrodes in six of the eight rats. The 50% point (the current level resulting in 50% detections) varied from 16.7 to 69.2 microA, with an overall mean of 42.4 (+/-8.1) microA across all single electrodes. Cortical micro-stimulation induced sensory-evoked behavior with similar characteristics as normal auditory stimuli. The results highlight the importance of the auditory cortex in a discrimination task and suggest that micro-stimulation of the auditory cortex might be an effective means for a graded information transfer of auditory information directly to the brain as part of a cortical auditory prosthesis.

A finite element model has been developed to investigate the theoretical relationship between changes in extracellular resistivity and electrical potential in a chronic extracellular recording scenario. The inputs to the model are experimental results obtained from chronic recording and complex impedance measurements in cerebral cortex of adult guinea pigs. Using the measured tissue-electrode impedance to set the resistivity in the model provides simulated extracellular potentials that are consistent with the measured spike amplitudes. In both the experimental and theoretical paradigms it was found that increased extracellular resistivity results in an increased potential at the recording electrode tip. Although the results of the two methods could not be directly correlated, they do suggest that a certain amount of variance can be accounted for by the increased resistivity values. The methods presented offer a powerful theoretical tool for understanding some of the factors, which may affect chronic extracellular recording stability.

A direct interface with motor cortical neurons could provide an optimal signal for controlling a neural prosthesis. In order to accomplish this, a real-time conversion of simultaneously recorded neural activity to an on-line command for movement is required. A system has been established to isolate the cellular activity of a group of cortical neurons and interpret their movement related information with a minimal delay. Data from a two-month simulation of this system is presented.

The cochlear nucleus (CN) transforms the spike-discharge activity of auditory-nerve fibers to form second-order sound representations. Onset units are one subclass of CN neurons and are notable for their response to changes in the synchrony of their high frequency auditory-nerve fiber inputs. The authors show that the mechanisms underlying this response can be used to inspire an algorithm for estimating the fundamental frequency of speech in background noise using standard Fourier-based methods.

The flow properties of liquid calcium alginate injections were investigated for application in endovascular embolization. Alginate shear properties were assessed with a rheometer and a controlled injection system. The experimental results were used to model the flow properties and predict alginate's flow characteristics within various medical microcatheter delivery systems. The results suggest that alginates undergo shear-thinning effects with increasing shear. A flow comparison of 2.0 wt % alginate and a Newtonian fluid (82 cP) injected from the same microcatheter had similar flow rates at low injection pressure (100 kPa). However, at high injection pressure (2100 kPa), the alginate was injectable at a flow rate 100% higher than was the Newtonian fluid. Further analysis of injections through microcatheters resulted in a flow model for predicting viscosity changes, flow rates, and injection pressures of liquid alginate at medium-to-high shear rates. The predicted injection pressures and flow rates had an average variance of less than 15% from that of the experimental flow data. This study indicates that calcium alginate has the requisite flow properties for successful delivery to vascular lesions via endovascular injection. Possible uses of alginates include treating arteriovenous malformations (AVMs), aneurysms, blood flow to tumors, and vascular hemorrhages.

Self-organizing map (SOM) has been applied in many different fields of science and engineering. In this chapter it is shown how the SOM can be used to decode neural spike trains of an awake animal and associate external auditory stimuli with spike patterns of the brain. This chapter begins with an introduction of the SOM and highlights advantages of the SOM when compared to other multivariate statistical data analysis tools commonly used to analyze similar data sets. In this study, simultaneous multichannel recording from guinea pig auditory cortex is examined using the SOM to assess the effectiveness and potential of the SOM in neurophysiological studies.

We sought to assess the stability of calcium alginate as an embolic agent in an animal model of a cerebral arteriovenous malformation (AVM). Swine cerebral AVM models were used to test the injectability, radiographic visualization, mechanical stability, and biocompatibility of calcium alginate as an occlusive agent.The swine cerebral AVM model included a carotid-to-jugular anastomosis to increase flow to the rete mirabile (RM), thereby simulating the pressure gradient and shunted blood flow of an AVM. Alginate and the reactive component, calcium chloride, were injected from double-lumen microcatheters to form a complete RM occlusion in the "acute swine" AVM model and a partial occlusion in a "survival swine" model.Angiographic and histological results verified complete occlusion of the left RM in acute animals when alginate was injected in stages. Partial RM occlusion in the survival animals blocked blood flow to the inferior portion of the RM but left flow open to the superior portion of the RM and the circle of Willis. One-week survival results showed that the alginate remained a stable occlusive material. Histological results showed a minor bioactive response and encapsulation of the alginate polymer, thereby increasing the stability and effective occlusion of the embolization material.Calcium alginate proved to be an effective endovascular occlusion material that blocked blood flow to the swine RM. The swine AVM models provided assessment of alginate injectability and effective occlusion and provided initial in vivo characteristics of alginate stability and biocompatibility.

This paper describes a detailed protocol for obtaining chronic, multi-site unit recordings in cerebral cortex of awake animals for periods of three months or more. The protocol includes details for making relatively simple and inexpensive implantable multichannel electrodes that consist of arrays of separate microwires. The results reported in this paper suggest that a viable implant will have discriminable unit activity on about 80% of the electrodes, resulting in, on average, the simultaneous unit recording of upwards of 60 units during a daily recording session. The active electrodes during one recording session tend to remain active in subsequent recording sessions for several weeks. Using the methods described here, implants have been constructed which incorporate several different electrode materials, coatings, sizes, and electrode separation within a single array. These microwire electrode arrays provide the basic technology for obtaining unit recordings for several months. This provides a model system for studying biocompatibility of neural implants, which is a critical component for the development of neural implants that have an indefinite working span. Themes: Sensory Systems Topics: Auditory systems: central physiology

Octopus cells are one of the principal cell types in the mammalian posteroventral cochlear nucleus. These cells respond to the onset of a toneburst with a precisely timed spike followed by little, if any, sustained activity. While experimental studies have partially characterized the cell, the mechanisms of this onset response are not well understood. The present study involved a model-based investigation that analyzed the responses of a compartmental model of the octopus cell in terms of synaptic effectiveness and dynamic spike threshold. The simulations demonstrate that properties of the onset response (first-spike latency, temporal precision of the first spike, and sustained firing rate) can be predicted from the values of these cell properties for a wide range of model configurations. These relationships were further analyzed through the development of mathematical expressions for synaptic effectiveness and dynamic spike threshold. This computational analysis resulted in a relatively simple explanation of the onset response, as well as predictions of the responses of octopus cells to nontonal, complex stimuli.

We developed a biologically plausible, compartmental model of stellate cells in the cochlear nucleus (CN) to investigate the effects of synaptic distribution on spike activity using constant and amplitude-modulated (AM) tones. Differences in the spatial distribution of excitatory auditory nerve (AN) inputs and inhibitory non-cochlear inputs evoke a range of post-stimulus time histogram (PSTH) chopper responses from sustained to transient to onset. These simulations illustrate how intrinsic membrane properties and non-linear dendritic processing of AN inputs determine the temporal responses characteristic of stellate cells observed in PSTHs and modulation transfer functions (MTFs).

The cochlear nucleus (CN) is the gateway to the central auditory system. The auditory nerve branches and terminates in the CN forming separate tonotopic maps in each of the nucleus’s three subdivisions. Octopus cells are found exclusively in the posteroventral cochlear nucleus and project to the inferior colliculus. These cells are sensitive to physical properties of sound, such as frequency, intensity, and amplitude fluctuations, although their role in auditory processing is unknown. Over a 40 dB dynamic range, the octopus cell always fires an action potential at the onset of a pure tone and then fires randomly at much lower rates throughout the rest of the tone. This onset response has not yet been explained in terms of the cell’s known physiological and anatomical properties, which include a fast membrane, highly convergent primary afferent inputs, and few inhibitory inputs. A better understanding of the properties underlying this response will provide increased, quantitative insight into auditory processing.

Brain–machine interface decoding algorithms need to be predicated on assumptions that are easily met outside of an experimental setting to enable a practical clinical device. Given present technological limitations, there is a need for decoding algorithms which (a) are not dependent upon a large number of neurons for control, (b) are adaptable to alternative sources of neuronal input such as local field potentials (LFPs), and (c) require only marginal training data for daily calibrations. Moreover, practical algorithms must recognize when the user is not intending to generate a control output and eliminate poor training data. In this paper, we introduce and evaluate a Bayesian maximum-likelihood estimation strategy to address the issues of isolating quality training data and self-paced control. Six animal subjects demonstrate that a multiple state classification task, loosely based on the standard center-out task, can be accomplished with fewer than five engaged neurons while requiring less than ten trials for algorithm training. In addition, untrained animals quickly obtained accurate device control, utilizing LFPs as well as neurons in cingulate cortex, two non-traditional neural inputs.

We investigated using poly(3,4-ethylenedioxythiophene) (PEDOT) to lower the impedance of small, gold recording electrodes with initial impedances outside of the effective recording range. Smaller electrode sites enable more densely packed arrays, increasing the number of input and output channels to and from the brain. Moreover, smaller electrode sizes promote smaller probe designs; decreasing the dimensions of the implanted probe has been demonstrated to decrease the inherent immune response, a known contributor to the failure of long-term implants. As expected, chronically implanted control electrodes were unable to record well-isolated unit activity, primarily as a result of a dramatically increased noise floor. Conversely, electrodes coated with PEDOT consistently recorded high-quality neural activity, and exhibited a much lower noise floor than controls. These results demonstrate that PEDOT coatings enable electrode designs 15 microns in diameter.

We describe a computational model of the Limulus lateral eye to investigate the relationship between the neural coding of the retina and the visual performance of the animal. The network model, implemented on the massively parallel Connection Machine, computes the firing rates of optic-nerve fibers for simulated visual scenes. Intensity-response functions computed for individual nerve fibers that project to the brain agree with experimental measurements. Responses computed for the entire array of retinal neurons show that moving targets enhance the modulation of optic-nerve activity.

Penetrating neural probe technologies allow investigators to record electrical signals in the brain. The implantation of probes causes acute tissue damage, partially due to vasculature disruption during probe implantation. This trauma can cause abnormal electrophysiological responses and temporary increases in neurotransmitter levels, and perpetuate chronic immune responses. A significant challenge for investigators is to examine neurovascular features below the surface of the brain in vivo. The objective of this study was to investigate localized bleeding resulting from inserting microscale neural probes into the cortex using two-photon microscopy (TPM) and to explore an approach to minimize blood vessel disruption through insertion methods and probe design. 3D TPM images of cortical neurovasculature were obtained from mice and used to select preferred insertion positions for probe insertion to reduce neurovasculature damage. There was an 82.8 +/- 14.3% reduction in neurovascular damage for probes inserted in regions devoid of major (5 microm) sub-surface vessels. Also, the deviation of surface vessels from the vector normal to the surface as a function of depth and vessel diameter was measured and characterized. 68% of the major vessels were found to deviate less than 49 microm from their surface origin up to a depth of 500 microm. Inserting probes more than 49 microm from major surface vessels can reduce the chances of severing major sub-surface neurovasculature without using TPM.

Object The availability of sophisticated neural probes is a key prerequisite in the development of future brain-machine interfaces (BMIs). In this study, the authors developed and validated a neural probe design capable of simultaneous drug delivery and electrophysiology recordings in vivo. Focal drug delivery promises to extend dramatically the recording lives of neural probes, a limiting factor to clinical adoption of BMI technology. Methods To form the multifunctional neural probe, the authors affixed a 16-channel microfabricated silicon electrode array to a fused silica catheter. Three experiments were conducted in rats to characterize the performance of the device. Experiment 1 examined cellular damage from probe insertion and the drug distribution in tissue. Experiment 2 measured the effects of saline infusions delivered through the probe on concurrent electrophysiological measurements. Experiment 3 demonstrated that a physiologically relevant amount of drug can be delivered in a controlled fashion. For these experiments, Hoechst and propidium iodide stains were used to assess insertion trauma and the tissue distribution of the infusate. Artificial CSF (aCSF) and tetrodotoxin (TTX) were injected to determine the efficacy of drug delivery. Results The newly developed multifunctional neural probes were successfully inserted into rat cortex and were able to deliver fluids and drugs that resulted in the expected electrophysiological and histological responses. The damage from insertion of the device into brain tissue was substantially less than the volume of drug dispersion in tissue. Electrophysiological activity, including both individual spikes as well as local field potentials, was successfully recorded with this device during real-time drug delivery. No significant changes were seen in response to delivery of aCSF as a control experiment, whereas delivery of TTX produced the expected result of suppressing all spiking activity in the vicinity of the catheter outlet. Conclusions Multifunctional neural probes such as the ones developed and validated within this study have great potential to help further understand the design space and criteria for the next generation of neural probe technology. By incorporating integrated drug delivery functionality into the probes, new treatment options for neurological disorders and regenerative neural interfaces using localized and feedback-controlled delivery of drugs can be realized in the near future.

Improving cortical prostheses requires the development of recording neural interfaces that are efficient in terms of providing maximal control information with minimal interface complexity. While the typical approaches have targeted neurons in the motor cortex with multiple penetrating shanks, an alternative approach is to determine an efficient distribution of electrode sites within the layers of the cortex with fewer penetrating shanks. The objective of this study was to compare unit activity in the upper and lower layers of the cortex with respect to movement and direction in order to inform the design of penetrating microelectrodes. Four rats were implanted bilaterally with multi-site single-shank silicon microelectrode arrays in the neck/shoulder region of the motor cortex. We simultaneously recorded unit activity across all layers of the motor cortex while the animal was engaged in a movement direction task. Localization of the electrode array within the different layers of the cortex was determined by histology. We denoted units from layers 2 and 3 and units as upper layer units, and units from layers 5 and 6 as lower layer units. Analysis of unit spiking activity demonstrated that both the upper and lower layers encode movement and direction information. Unit responses in either cortical layer of the cortex were not preferentially associated with contralateral or ipsilateral movement. Aggregate analysis (633 neurons) and best session analysis (75 neurons) indicated that units in the lower layers (layers 5, 6) are more likely to encode direction information when compared to units in the upper layers (layers 2, 3) (p< 0.05). These results suggest that electrode sites clustered in the lower layers provide access to more salient control information for cortical neuroprostheses.

The neocortex is a six-layered tissue consisting of different cell types. How does unit activity in the different layers of the motor cortex relate to movement? Does implantation in a particular layer improve direction decoding ability for a neuroprosthetic device? We simultaneously recorded unit activity in different layers of the rat motor cortex using chronic multi-site silicon electrodes. We used a combination of histology and electrophysiological signatures of Local Field Potentials (LFPs) to accurately localize the electrode sites in the different layers of the cortex. We analyzed 142 units from two animals and found that 40 units (28%) in Layers II to V showed significant modulation with respect to movement. Of these units that showed significant modulation, 9/20 (45%) of units in Layers II/III encoded directional information as compared to 15/19 (79%) of the units in Layers IV/V. These preliminary results suggest that units in Layers IV/V relatively contain more directional information than other layers of the cortex.