Ethan D Cohen

Prosthetic interfaces with the visual system: biological issues

The design of effective visual prostheses for the blind represents a challenge for biomedical engineers and neuroscientists. Significant progress has been made in the miniaturization and processing power of prosthesis electronics; however development lags in the design and construction of effective machine-brain interfaces with visual system neurons. This review summarizes what has been learned about stimulating neurons in the human and primate retina, lateral geniculate nucleus and visual cortex. Each level of the visual system presents unique challenges for neural interface design. Blind patients with the retinal degenerative disease retinitis pigmentosa (RP) are a common population in clinical trials of visual prostheses. The visual performance abilities of normals and RP patients are compared. To generate pattern vision in blind patients, the visual prosthetic interface must effectively stimulate the retinotopically organized neurons in the central visual field to elicit patterned visual percepts. The development of more biologically compatible methods of stimulating visual system neurons is critical to the development of finer spatial percepts. Prosthesis electrode arrays need to adapt to different optimal stimulus locations, stimulus patterns, and patient disease states.

MIDA: A Multimodal Imaging-Based Detailed Anatomical Model of the Human Head and Neck

Computational modeling and simulations are increasingly being used to complement experimental testing for analysis of safety and efficacy of medical devices. Multiple voxel- and surface-based whole- and partial-body models have been proposed in the literature, typically with spatial resolution in the range of 1–2 mm and with 10–50 different tissue types resolved. We have developed a multimodal imaging-based detailed anatomical model of the human head and neck, named “MIDA”. The model was obtained by integrating three different magnetic resonance imaging (MRI) modalities, the parameters of which were tailored to enhance the signals of specific tissues: i) structural T1- and T2-weighted MRIs; a specific heavily T2-weighted MRI slab with high nerve contrast optimized to enhance the structures of the ear and eye; ii) magnetic resonance angiography (MRA) data to image the vasculature, and iii) diffusion tensor imaging (DTI) to obtain information on anisotropy and fiber orientation. The unique multimodal high-resolution approach allowed resolving 153 structures, including several distinct muscles, bones and skull layers, arteries and veins, nerves, as well as salivary glands. The model offers also a detailed characterization of eyes, ears, and deep brain structures. A special automatic atlas-based segmentation procedure was adopted to include a detailed map of the nuclei of the thalamus and midbrain into the head model. The suitability of the model to simulations involving different numerical methods, discretization approaches, as well as DTI-based tensorial electrical conductivity, was examined in a case-study, in which the electric field was generated by transcranial alternating current stimulation. The voxel- and the surface-based versions of the models are freely available to the scientific community.

Safety and effectiveness considerations for clinical studies of visual prosthetic devices

With the advent of new designs of visual prostheses for the blind, FDA is faced with developing guidance for evaluating their engineering, safety and patient performance. Visual prostheses are considered significant risk medical devices, and their use in human clinical trials must be approved by FDA under an investigation device exemption (IDE). This paper contains a series of test topics and design issues that sponsors should consider in order to assess the safety and efficacy of their device. The IDE application includes a series of pre-clinical and clinical data sections. The pre-clinical section documents laboratory, animal and bench top performance tests of visual prostheses safety and reliability to support a human clinical trial. The materials used in constructing the implant should be biocompatible, sterile, corrosion resistant, and able to withstand any forces exerted on it during normal patient use. The clinical data section is composed of items related to patient-related evaluation of device performance. This section documents the implantation procedure, trial design, statistical analysis and how visual performance is assessed. Similar to cochlear implants, a visual prosthesis is expected to last in the body for many years, and good pre-clinical and clinical testing will help ensure its safety, durability and effectiveness.

Effects of high-level pulse train stimulation on retinal function

We examined how stimulation of the local retina by high-level current pulse trains affected the light-evoked responses of the retinal ganglion cells. The spikes of retinal ganglion cell axons were recorded extracellularly using an in vitro eyecup preparation of the rabbit retina. Epiretinal electrical stimulation was delivered via a 500 microm inner diameter saline-filled, transparent tube positioned over the retinal surface forming the receptive field center. Spot stimuli were presented periodically to the receptive field center during the experiment. Trains of biphasic 1 ms current pulses were delivered to the retina at 50 Hz for 1 min. Pulse train charge densities of 1.3-442 microC/cm(2)/phase were examined. After pulse train stimulation with currents >or=300 microA (133 microC/cm(2)/phase), the ganglion cells ability to respond to light was depressed and a significant time was required for recovery of the light-evoked response. During train stimulation, the ganglion cells ability to spike following each current pulse fatigued. The current levels evoking train-evoked depression were suprathreshold to those evoking action potentials. Train-evoked depression was stronger touching the retinal surface, and in some cases impaired ganglion cell function for up to 30 min. This overstimulation could cause a transient refractory period for electrically stimulated perception in the retinal region below the electrode.

Light-Induced Thickening of Photoreceptor Outer Segment Layer Detected by Ultra-High Resolution OCT Imaging.

Purpose We examined if light induces changes in the retinal structure that can be observed using optical coherence tomography (OCT). Methods Normal C57BL/6J mice (age 3–6 months) adapted to either room light (15 minutes to ∼5 hours, 50–500 lux) or darkness (overnight) were imaged using a Bioptigen UHR-OCT system. Confocal histologic images were obtained from mice killed under light- or dark-adapted conditions. Results The OCT image of eyes adapted to room light exhibited significant increases (6.1 ± 0.8 μm, n = 13) in total retina thickness compared to the same eyes after overnight dark adaptation. These light-adapted retinal thickness changes occurred mainly in the outer retina, with the development of a hyporeflective band between the RPE and photoreceptor-tip layers. Histologic analysis revealed a light-evoked elongation between the outer limiting membrane and Bruchs membrane from 45.8 ± 1.7 μm in the dark (n = 5) to 52.1 ± 3.7 μm (n = 5) in the light. Light-adapted retinas showed an increase of actin staining in RPE apical microvilli at the same location as the hyporeflective band observed in OCT images. Elongation of the outer retina could be detected even with brief light exposures, increasing 2.1 ± 0.3 μm after 15 minutes (n = 9), and 4.1 ± 1.0 μm after 2 hours (n = 6). Conversely, dark-adaptation caused outer retinal shortening of 1.4 ± 0.4 μm (n = 7) and 3.0 ± 0.5 μm (n = 8) after 15 minutes and 2 hours, respectively. Conclusions Light-adaption induces an increase in the thickness of the outer retina and the appearance of a hyporeflective band in the OCT image. This is consistent with previous reports of light-induced fluid accumulation in the subretinal space.

A novel high electrode count spike recording array using an 81,920 pixel transimpedance amplifier-based imaging chip

Microelectrode recording arrays of 60-100 electrodes are commonly used to record neuronal biopotentials, and these have aided our understanding of brain function, development and pathology. However, higher density microelectrode recording arrays of larger area are needed to study neuronal function over broader brain regions such as in cerebral cortex or hippocampal slices. Here, we present a novel design of a high electrode count picocurrent imaging array (PIA), based on an 81,920 pixel Indigo ISC9809 readout integrated circuit camera chip. While originally developed for interfacing to infrared photodetector arrays, we have adapted the chip for neuron recording by bonding it to microwire glass resulting in an array with an inter-electrode pixel spacing of 30 μm. In a high density electrode array, the ability to selectively record neural regions at high speed and with good signal to noise ratio are both functionally important. A critical feature of our PIA is that each pixel contains a dedicated low noise transimpedance amplifier (∼0.32 pA rms) which allows recording high signal to noise ratio biocurrents comparable to single electrode voltage amplifier recordings. Using selective sampling of 256 pixel subarray regions, we recorded the extracellular biocurrents of rabbit retinal ganglion cell spikes at sampling rates up to 7.2 kHz. Full array local electroretinogram currents could also be recorded at frame rates up to 100 Hz. A PIA with a full complement of 4 readout circuits would span 1cm and could acquire simultaneous data from selected regions of 1024 electrodes at sampling rates up to 9.3 kHz.

A computational model for bipolar deep brain stimulation of the subthalamic nucleus

Deep brain stimulation (DBS) of the subthalamic nucleus (STN) has been shown to reduce some of the symptoms of advanced, levodopa-responsive Parkinsons disease that are not adequately controlled with medication. However, the precise mechanism of the therapeutic action of DBS is still unclear. Stimulation-induced side effects are not uncommon and require electrical “dose” adjustments. Quantitative methods are needed to fully characterize the electric field in the deep brain region that surrounds the electrodes in order to help with adjustments and maximize the efficacy of the device. Herein we report a magnetic resonance imaging (MRI)-based head model proposed for analysis of fields generated by deep brain stimulation (DBS). The model was derived from multimodal image data at 0.5mm isotropic spatial resolution and distinguishes 142 anatomical structures, including the basal ganglia and 38 nuclei of the thalamus. Six bipolar electrode configurations (1-2, 1-3, 1-4, 2-3, 2-4, 3-4) were modeled in order to assess the effects of the inter-electrode distance of the electric field. Increasing the distance between the electrodes results in an attenuated stimulation, with up to 25% reduction in electric field amplitude delivered (2-3 vs. 1-4). The map of the deep brain structures provided a highly precise anatomical detail which is useful for the quantitative assessment of current spread around the electrode and a better evaluation of the stimulation setting for the treatment optimization.

The use of time-lapse optical coherence tomography to image the effects of microapplied toxins on the retina.

PURPOSE We developed a novel technique for accelerated drug screening and retinotoxin characterization using time-lapse optical coherence tomography (OCT) and a drug microapplication device. METHODS Using an ex vivo rabbit eyecup preparation, we studied retinotoxin effects in real-time by microperfusing small retinal areas under a transparent fluoropolymer tube. Known retinotoxic agents were applied to the retina for 5-minute periods, while changes in retinal structure, thickness, and reflectance were monitored with OCT. The OCT images of two agents with dissimilar mechanisms, cyanide and kainic acid, were compared to their structural changes seen histologically. RESULTS We found the actions of retinotoxic agents tested could be classified broadly into two distinct types: (1) agents that induce neuronal depolarization, such as kainic acid, causing increases in OCT reflectivity or thickness of the inner plexiform and nuclear layers, and decreased reflectivity of the outer retina; and (2) agents that disrupt mitochondrial function, such as cyanide, causing outer retinal structural changes as evidenced by a reduction in the OCT reflectivity of the photoreceptor outer segment and pigment epithelium layers. CONCLUSIONS Retinotoxin-induced changes in retinal layer reflectivity and thickness under the microperfusion tube in OCT images closely matched the histological evidence of retinal injury. Time-lapse OCT imaging of the microperfused local retina has the potential to accelerate drug retinotoxicological screening and expand the use of OCT as an evaluation tool for preclinical animal testing.

Access resistance of stimulation electrodes as a function of electrode proximity to the retina

OBJECTIVE Epiretinal prostheses seek to effectively stimulate the retina by positioning electrode arrays close to its surface so current pulses generate narrow retinal electric fields. Our objective was to evaluate the use of the electrical impedance of insulated platinum electrodes as a measure of the proximity of insulated platinum electrodes to the inner surface of the retina. APPROACH We examined the impedance of platinum disk electrodes, 0.25 mm in diameter, insulated with two widths (0.8 and 1.6 mm outer diameter) of transparent fluoropolymer in a rabbit retinal eyecup preparation. Optical coherence tomography measured the electrodes proximity to the retinal surface which was correlated with changes in the voltage waveform at the electrode. Electrode impedance changes during retinal deformation were also studied. MAIN RESULTS When the 1.6 mm diameter insulated electrodes advanced towards the retinal surface from 1000 μm, their voltage step at current pulse onset increased, reflecting an access resistance increase of 3880 ± 630 Ω, with the 50% midpoint averaging 30 μm, while thin 0.8 mm insulated electrode advancement showed an access resistance increase 50% midpoint averaging 16 μm. Using impedance spectroscopy, electrode-retina proximity differences were seen in the 1.6 mm insulated electrode impedance modulus between 1 and 100 kHz and the waveform phase angle at 0.3-10 kHz, while thin 0.8 mm insulated electrode advancement produced smaller impedance modulus changes with retinal proximity between 3 and 100 kHz. These impedance changes with retinal proximity may reflect different sized zones of eye wall being coupled in series with the insulated platinum electrode. SIGNIFICANCE The proximity of stimulus electrodes to neural tissue in fluid-filled spaces can be estimated from access resistance changes in the stimulus pulse waveform. Because many prosthetic devices allow back telemetry communication of the stimulus electrode waveform, it is possible these series resistance increases observed with retinal proximity could be used as a metric of stimulus electrode placement.

Computational platform combining detailed and precise functionalized anatomical phantoms with EM-Neuron interaction modeling

A computational framework including functionalized anatomical phantoms able to simulate electromagnetic (EM)-neuron interactions in complex tissue-structure environments was developed. The anatomic head model distinguishes a large number of structures, particularly in regions relevant to EM-neuron interactions (i.e., ear, eye, deep brain structures). Multimodal Magnetic Resonance (MR) images were acquired together with Diffusion Tensor Imaging (DTI) data to guide neuron model placement and inform about tissue anisotropy. A dedicated EM solver was implemented and coupled with dynamic models of neuronal activity. The topologically conforming, non-self-intersecting, high-element-quality surfaces are suitable for a wide range of numerical methods and solvers, as demonstrated in an application derived from transcranial alternating current stimulation. The platform was validated against literature data, e.g., on the SENN [1] model which was further extended to account for local thermal effects. Additionally, the EM-neuron coupling simulation platform was also applied to investigate MRI gradient coil switching induced nerve stimulation.