Challenges in an animal model for a Neuroprosthesis

by user


nail polish






Challenges in an animal model for a Neuroprosthesis
Challenges in an animal model for a Neuroprosthesis
Olaf Christ*, Ulrich G. Hofmann
Research Group for Neuroelectronic Systems, Department of Neurosurgery, University Medical Center Freiburg, Germany
[email protected]
Neuroprosthesis, animal model, microelectrodes, gliosis, artefacts.
On our ongoing journey on developing the ideal neuroprosthesis, we have devised an animal model
for in vivo testing. Our ultimate goal is developing a long lasting bidirectional neuroprosthesis using
a large array of recording sites. Naturally, we are faced with a plethora of challenges. To begin with,
doing long term chronic multisite recordings of neuron populations with a high number of channels,
gliosis will eventually increasingly jeopardize the recording quality over time. A possible solution to
this problem seems to be cellular modulation of polymeric device surfaces by using adult stem cells.
For closed-loop in vivo testing we propose a simple surrogate neuroprosthesis consisting of either a
physical or a computer simulated robot and a rat constantly exploring either a real exploratory maze
or a virtual environment. For closed-loop control we use a video tracking system for providing
ground truth position data to enable real time qualitative analysis of position information decoded
from neuronal signals. This is another big challenge, since we have to correctly understand and
decode the rat's internal allocentric map and at the same time understand how a rat uses this
information to perform self localization and navigation. For that, we use position information from
hippocampal place cells within CA1 and CA3 and localized firing patterns of grid cells. Cells which
are projection neurons representing geometric coordinates within an internal grid used for
navigation. Grid cells are found in the medial entorhinal cortex (MEC) and the pre- and
parasubiculum. Place cells encode areas and together with information from grid cells path
integration also known as simultaneous localization and mapping or SLAM is performed. That is,
grid cells remap position information from hippocampal place cells to enable the rat to navigate in a
known and unknown environment. Similar navigation systems are known to exist for a long time and
can be found in humans as well as in insects. Compared to CA1 and CA3, recordings are
considerably more difficult for grid cells, because the areas where grid cells are located are
significantly smaller. Our proposed solution for providing the high accuracy needed for implanting
electrodes is our custom built computer controlled robot for stereotactic surgery. In addition to
position information, we further extend our model by recording from the barrel cortex for detecting
information provided by the rat's whiskers and stimulation of the somatosensory cortex to induce the
sensation of touching an object with the whiskers. Together with stimulation of the medial forebrain,
it's even possible to remote control the rat. Now, if we employ simultaneous stimulation and
recording, vital for true bidirectional neuroprosthesis, we have to either algorithmically get rid of
possible artifacts or make sure that stimulation remains restricted to a small area and not directly
interfere with the distant or nearby recording area. A modification of the describe model is replacing
the physical exploratory maze with a virtual environment. By keeping the rat stationary and having
the rat facing a projection screen showing a computer generated virtual maze while walking on a 3D
treadmill e.g. a large ball, we can further isolate specific parameters, such as computer controlled
mechanical whisker stimulation or exposing the rat to a specific smell. Furthermore, we can outfit
the robot with artificial whiskers and use this sensory information for remote controlling the rat as
explained previously. Note, that being able to record and decode neuronal information and at the
same time doing neurostimulation is the essential challenge for a bidirectional neuroprosthesis.
Because of it's modular nature, our proposed animal model for a neuroprosthesis also provides some
usefulness for conducting behavioural studies.
Wireless Electrocorticograph (ECoG) Recording System
K. Shane Guillory*, Robert Askin, Christopher Smith, Daniel McDonnall
Ripple, Salt Lake City, USA
[email protected]
Electrocorticograph, wireless, cortical recording, brain mapping, brain-machine interface
Epilepsy is the second most common serious neurological condition in the United States after stroke.
Some of the patients that do not respond adequately to drug therapies are candidates for surgical
treatment to resect or disrupt seizure foci. The use of percutaneous leads from subdural electrodes
exposes patients to risks of infection from lead wires passing through the scalp. A wireless system
could eliminate the risk that patients develop infections at the site of percutaneous cables and allow
patients to be monitored outside of costly Neurology ICUs.
We are developing a wireless ECoG recording implant with a novel thin-film array connected to an
integrated telemetry module that is placed subcutaneously near the craniotomy. The implant will be
used with an external transceiver placed directly over the telemetry module of the implant and
incorporated in the head dressings of the patient. The transceiver can be connected to a bedside data
acquisition system for monitoring, or connected to a belt-worn recording unit for ambulatory and
outpatient use. The implant electronics will amplify, filter, and digitize the ECoG signals from the
electrode array. The external transceiver will inductively power the implant and receive the
transcutaneous infrared data that is sent back from the implant. This close-coupled inductive/optical
configuration for the telemetry is very immune to electromagnetic interference and susceptibility
We have developed prototype devices to test a wide variety of inductive coil configurations for
driving the coils in high-efficiency modes, as well as large prototype boards for the implant to test a
variety of power recovery methods. The implant and external transceiver prototypes provided robust
power transfer with telemetry coil distances up to 5 cm and alignment offsets up to 2 cm. The
transceivers were also tolerant of tilts up to 30 degrees. Overall power consumption for the system
was under 300mW for most alignments. Power efficiency will be optimized in subsequent iterations
of the system with new coil geometries. The prototype IR data transmission system was validated
with a 2cm thick sample of pig skin, twice the design limit of 1cm. These prototypes can transmit up
to 10 Megabits/sec of data with only 2mW of power to the infrared emitter and lateral external
receiver displacements up to 1 cm.
To keep the device cost low, we have developed an integrated process for creating the electrode
array and encapsulating the implant electronics with patterned layers of insulating and conductive
polymers. The conductive polymers are composed of polymer filled with conductive metal
microparticles, and these are deposited to form surface electrode sites and internal electrical traces
and leads. We have developed processes for fabricating these electrode arrays using both silicone
and urethane polymers, with silver, gold, and platinum conductive particles. These processes allow
production of composite structures as thin as 250 μm, with conductive features as small as 150 μm,
and trace and space pitches down to 300 μm. The traces can maintain resistances of less than 100
ohms/cm and tolerate stretch deformations up to 10%.
We have tested combinations of implant materials with accelerated lifetime testing protocols (67°C,
10V bias) to simulate implant periods up to one year. We have also constructed assemblies for
accelerated flexural lifetime testing, and have verified that prototype devices maintain trace
conductivity and insulation layer integrity for over 800,000 flex cycles. We are now beginning in
vivo testing of these polymer electrodes and integration of the implant electronics.
This work is funded by a National Institutes of Health SBIR Fast-Track grant (R44 NS061604).
Online Classification of code-based VEPs using ECoG
C. Kapeller1*, R. Prückl1, E. Pataraia2, G. Lindinger2, T. Czech2, C. Guger1
Guger Technologies OG, Schiedlberg, Austria
Medical University of Vienna, Vienna, Austria
[email protected]
Keywords (up to 5)
A brain-computer-interface (BCI) allows the user to control a device like e.g. a neuroprosthesis with
brain activity [1]. Usually, BCIs use brain activity extracted from the electroencephalogram (EEG)
or the electrocorticogram (ECoG) [2]. Some of them rely on visual stimuli that elicit steady-state
visual evoked potential (SSVEP) [3] or code-based VEPs [4]. This work investigates the comparison
of code-based VEPs in the EEG and ECoG, as well as the online classification performance of a
corresponding BCI system. The BCI system is based on MATLAB/Simulink and performs online
and offline signal processing. Online classification results can be sent an arbitrary device. For visual
stimulation a 63 bit pseudo-random m-sequence was presented on a standard 60 Hz LCD monitor.
One subject, who was suffering from epilepsy and therefore having six intracranial electrode grids
across the left hemisphere (Figure 1, left) participated in the ECoG recordings, where another healthy
subject provided comparable EEG data. Data was recorded using a bio-signal amplifier with a
sampling rate of 256 Hz and band-pass filtered between 0.5 and 30 Hz. As the visual cortex was not
covered by all electrodes, only electrodes that were closest to the visual cortex were used for the
analysis. The first task (i) was to gaze at a reference target over 200 cycles, to extract an average
template signal from the ECoG for each electrode, with respect to the sequence (Figure 2). Figure 3
shows that ECoG based templates (up to 50 µV for E1) have higher response to the visual
stimulation compared to EEG templates (up to 5 µV for O1). In a second task (ii), four target
sequences were presented on the screen simultaneously. Each sequence is a phase shifted version of
the reference sequence from (i). The subject had to gaze at each target three times for 10 s (3 s of rest
and 7 s flickering), which led to 12 trials in total. For online target identification, a 2.1 s (two
sequence cycles) long signal buffer is compared with the previously calculated templates using a
canonical correlation analysis (CCA). Therefore, the canonical vector based on the templates and the
raw data, is calculated offline and then used as a spatial filter, to combine the signal channels in the
online experiment. The correlation coefficients between the spatial filtered signal and the phase
shifted templates are the features in a linear discriminant analysis (LDA). The LDA classifier is
computed offline, based on the comparison of the raw data from (i) and four phase shifted versions
of the templates and then applied for online classification. After finishing (ii), the ECoG experiment
showed a maximum online classification accuracy of 83 % (Figure 1, right). In another study using
EEG, the mean online classification accuracy of eleven subjects using the same system was 93.66 %.
Nevertheless, the coverage of the visual cortex was much better in the EEG experiments compared to
the ECoG subject. A visual inspection of the ECoG templates shows that only electrode E1 has a
good response to the visual stimulation. The presented system can be used for continuous control of
a device, as well as trial based control for high-level commands. This provides a flexible
environment to control external devices like neuroprosthesis. The current implementation contains a
latency of about 2 s, which influences the reaction time, especially in the continuous mode. As ECoG
provides higher temporal resolution compared to EEG, higher stimulation frequencies in
combination with a smaller signal buffer might be useful, to reduce the latency.
Figure 1: Electrode montage with 6 strips (A1-A12, B1-B10, C1-C12, D1-D8, E1-E10, and F1-F10) across the left hemisphere
(left). Online classification error of code-based VEPs, based on four phase shifted visual stimulation sequences (right).
Figure 2: Templates of the individual electrode positions according to the ECoG grid E and F of the subject. Only the first four
electrodes of the grids were used, as they cover best the visual cortex. Electrode E1 shows the highest amplitudes in the VEPs.
Figure 3: Comparison of an ECoG template (right) and an EEG template (left). The EEG signal was recorded from a different
subject. The O1 position was chosen for comparsion, as it overlaps best with the E1 position of the ECoG recordings.
J.R. Wolpaw, N. Birbaumer, D.J. McFarland, G. Pfurtscheller, and T.M. Vaughan, “Brain-computer interfaces for
communication and control,” Clinical neurophysiology : official journal of the International Federation of
Clinical Neurophysiology, vol. 113, (no. 6), pp. 767-91, Jun 2002.
E.C. Leuthardt, G. Schalk, J.R. Wolpaw, J.G. Ojemann, and D.W. Moran, “A brain-computer interface using
electrocorticographic signals in humans,” Journal of neural engineering, vol. 1, (no. 2), pp. 63-71, Jun 2004.
Y. Wang, X. Gao, B. Hong, C. Jia, and S. Gao, “Brain-computer interfaces based on visual evoked potentials,”
IEEE engineering in medicine and biology magazine : the quarterly magazine of the Engineering in Medicine &
Biology Society, vol. 27, (no. 5), pp. 64-71, Sep-Oct 2008.
G. Bin, X. Gao, Y. Wang, Y. Li, B. Hong, and S. Gao, “A high-speed BCI based on code modulation VEP,” Journal
of neural engineering, vol. 8, (no. 2), pp. 025015, Apr 2011.
A Wireless Bidirectional Brain-Machine Interface
Suitable for Long-Term Implantation
*Fabian Kohler1, Martin Schuettler1,2, Joerg Fischer2, Rainer Mohrlok3, Jens Paetzold3,
Karl-Heinz Boven3, Andreas Moeller3, Christian Henle2, Wolfgang Meier2, Markus Raab2,
Thomas Fehrenbacher2, Juan S. Ordonez1, Xi Wang1,4, Tonio Ball4,5, Thomas Stieglitz1,2,5,
Joern Rickert2,5
Laboratory for Biomedical Microtechnology, Dept. of Microsystems Engineering, University of Freiburg, Germany
CorTec GmbH, Freiburg, Germany 3 Multi Channel Systems GmbH, Reutlingen, Germany
University Hospital, Freiburg, Germany 5 Bernstein Center Freiburg, Germany
*email: [email protected]
bci, neural interface, implant, recording, stimulation
Aiming at a chronically implantable brain-machine interface for cortical recording and stimulation in
humans, we developed a technological platform called BrainCon that has the potential of rapid
transfer to clinical trials and CE-market approval. Currently, our first implant version records from
16 channels (1 kHz sampling rate, 16 bit dynamic range) and allows for electrical stimulation (17 V
compliance voltage) on 8 channels. It is inductively powered using a carrier frequency of 16 MHz,
allowing a separation of 20 mm between extra corporal transmitter and implant. Data is transferred
via an infrared optical half-duplex link at a data rate of 1 Mbit/s. The implanted electronics are
hermetically sealed in a custom designed ceramic package, which consists of a screen printed base
substrate (635 µm, 96% Al2O3) and a screen printed lid substrate soldered to it with the electronics
placed in between [1]. The package lid permits a sufficient part of the infrared light to pass, allowing
the optical data transmitter to be placed inside the package. 50 cm long wires made from ∅ = 70 µm
MP35N with polyesterimide insulation are bundled in a silicone hose of 1.8 mm outer diameter. One
end of the wires is welded to platinum platelets, which are soldered to the screen printed
feedthroughs of the package base substrate, before the package is casted in medical grade silicone
adhesive. The 0.4 mm thick laser-fabricated [2] electrode array has 4 x 8 platinum/iridium contacts
(∅ = 0.8 mm, 4 mm center-to-center distance) and two additional rows of 7 contacts serving as
reference, embedded in a 19 x 33 mm² parylene C reinforced silicone rubber substrate. The wires are
resistance-welded to the tracks integrated in the electrode array and the welds are sealed with
silicone adhesive. The power receiving coil (outside the package) carries a permanent magnet in its
center which eases the alignment of the extra corporal transmitter coil through the skin after
implantation. Figure 1 shows a photograph of the assembled BrainCon implant. For operation of the
implant, the extra-corporal unit (not shown) has to be placed on the skin, aligned to implant. The unit
generates the magnetic field for implant powering and communicates with the implant via infrared
light. A USB 2.0 port of a personal computer is utilized for supplying the external unit with power
and for communicating with it. The personal computer runs custom-built software, allowing the user
to set the implant's stimulation paradigm via a graphical user interface. The software also receives
the stream of data from the implant's 16 recording channels, provides various data processing tools
like feature extraction and stores the recorded data to a hard drive.
Figure 1: Photo of BrainCon implant. A: Laser-fabricated electrode array, B: Electrode cable,
C: Hermetic electronic package, casted in medical grade silicone rubber, D: Dacron-mesh reinforced
rubber flaps for suture fixation, E: Radio frequency receiver coil and alignment magnet, rubber
Schuettler, M., Kohler, F., Ordonez, J. S. Stieglitz, T.: Hermetic Electronic Packaging of an
Implantable Brain-Machine-Interface with Transcutaneous Optical Data Communication.
Proceedings to the Annual Conference of IEEE EMBS, pp. 3886-3889, (2012).
M. Schuettler, S. Stiess, B. King, G.J. Suaning: Fabrication of Implantable Microelectrode
Arrays by Laser-Cutting of Silicone Rubber and Platinum Foil. Journal of Neural Engineering,
No. 2, pp. 121-128, doi:10.1088/1741-2560/2/1/013, (2005).
Enhanced Peri-Event Time Histograms from Intracortical Recordings
with Matched Wavelets for Spike Detection
Vahid Shalchyan1, 2, Sofyan Hammad1, Winnie Jensen1, Dario Farina2*
Dept. Health Science and Technology, Aalborg University, Aalborg, Denmark
Dept. of Neurorehabilitation Engineering, University Medical Center Göttingen, Göttingen, Germany
*[email protected]
Spike detection, wavelet transform, optimization, peri-event time histogram.
Accurate detection of action potentials is a basic step for decoding the embedded information in a
neural recording. Intra-cortical electrodes usually record the extracellular activity of single or
multiple neuronal units as well as the background of physical and biological noise. Several methods
have been proposed to discriminate the neural spikes from the background noise, ranging from
amplitude thresholding to template matching, multi-resolution operators, and wavelet-based
detectors. The detection performance of time-domain thresholding methods degrade in case of low
signal-to-noise ratio (SNR). Other methods show good detection performances when the shape or
time-frequency characteristics of the existing spike waveforms in the recording are known a priori.
The objectives of this study were: 1) to develop an optimal method for the neural spike detection
automatically without any prior assumption about the spike characteristics; and 2) to study the
impact of the proposed method on the extraction of the neural correlates of movement in a behavioral
experiment. For the first objective, we introduced a wavelet-based framework for the spike detection
which automatically finds the matched mother wavelet from a large dictionary of wavelet filters. The
proposed method was tested and compared with other matched wavelet approaches, previously
proposed, by using simulated data. For the second objective, the proposed method was compared
with other methods by studying their effects on the resulting peri-event time histogram (PETH) in a
behavioral experiment on intra-cortically implanted rats.
We generated a large dictionary of wavelet filters by scaling a group of compact support, orthogonal
and biorthogonal wavelet basis functions (Haar, Daubechies2, Symlet4, Coiflet1, Biorthogonal1.3,
Biorthogonal1.5, and Reverse Biorthogonal2.2 [1]). The scaling resolution was set 3 times finer than
dyadic scaling. The projections of the signal on the wavelet filters were ranked by a “spikiness”
score which was defined as estimated number of containing spikes. Three optimal matched filter
projections were combined by summing the absolute values of the large coefficients (i.e., above a
certain threshold) to build a new manifestation variable for spike detection as described in [2].
Using a library of spike templates from experimental recordings, 50 generations of three-unit spike
patterns were made. The generated patterns were shifted over the frequency scale by band-pass
filtering to uniformly cover the frequency range of 300-3000Hz. The background noises were
simulated as colored noise by an auto-regressive (AR) model of experimental noise recordings. The
simulated signals were analyzed with 5 levels of added noise, corresponding to SNRs from 1 to 1.5,
with increments of 0.1. Five male, Sprauge-Dawley rats were trained to perform the task of hitting a
paddle key three times consecutively with the forepaw of the preferred hand to get food rewards. The
area related to forelimb movement in the primary motor cortex (M1) was implanted by an array of
16-channel tungsten wires. The digital signal indicating the paddle hit responses and the control
signals were recorded along with the neural data for the analysis [3]. The experiment was performed
for every rat in few recording sessions in which the rat repeats the trained behavioral task over many
trials. For this study we used the data from four rats (two sessions for each rat). For each examined
recording session, the neural activities of the successful trials were extracted. An electrical
microstimulation was used to demonstrate causal links between multi-units neural activities for each
Table 1. Comparison of relative enhancements (%) in MRNP
for CWT and FDWT with respect to THR over all recording
channels for each rat in the experimental dataset.
( Mean SE)
( Mean SE)
Rat I
Rat II
Rat IV
Fig. 1. Average AUC measures over all simulated data as function
of SNR. The FDWT is compared to CWT and THR methods.
of recording channel and behavioral functions. Only the channels that showed hand movement
functions were included in this study. In each recording session, for each channel, all trials were
centered relative to the time of the first paddle hit. The time period was divided into a series of timebins (20 ms) and the average PETH was calculated over trials. To evaluate the ability of the methods
to predict the movement over trials, a measure of movement-related neural prediction (MRNP) was
defined as relative peak activity of PETH divided by the standard deviation of the background
activity in the PETH.
The compared methods in the simulation study include the proposed fine discrete wavelet transform
(FDWT), the continuous wavelet transform (CWT) proposed in [4], and the amplitude thresholding
(THR) method, as described in [5]. For the CWT method, the code from the author’s website was
used with 6 scales and Biorthogonal1.3 wavelet basis which had optimal results over all possible
wavelet functions. The receiver operator characteristic (ROC) curves and the area under the curves
(AUC) were used to evaluate the detection performance in the simulations. In the experimental
study, we reported the relative percentage of enhancement in the MRNP for CWT and FDWT with
respect to THR method averaged over channels and sessions for each rat.
Fig. 1 shows the average detection results in terms of AUC measures over all simulated datasets as a
function of SNR for different methods. The FDWT proposed method outperformed the THR and
CWT methods in all cases for the simulated dataset. Table 1 reports the average results of relative
enhancements in MRNP over all recording channels for each rat. The results indicate that the FWDT
outperformed the CWT in terms of enhancement in the movement-related neural prediction with
respect to the THR method. In conclusion, a method is proposed for single channel spike detection
which automatically finds the matched wavelet projections of the signal without user intervention.
The proposed method enhanced the prediction of movement from neural activity over trials by
enhancing the movement-related peak of the PETH.
Daubechies, I. (1992). Ten lectures on wavelets. Society for industrial and applied mathematics.
Shalchyan, V., Jensen, W., & Farina, D. (2012). Spike Detection and Clustering With Unsupervised Wavelet
Optimization in Extracellular Neural Recordings. Biomedical Engineering, IEEE Transactions on, 59(9), 2576-2585.
[3] Hammad, S., Kamavuako, E.N., Farina, D., & Jensen, W. (2012). Classification of a self-paced hitting task in freelymoving rats: Influence of denoising. (Manuscript in preparation)
[4] Nenadic, Z., & Burdick, J. W. (2005). Spike detection using the continuous wavelet transform. Biomedical
Engineering, IEEE Transactions on, 52(1), 74-87.
[5] Quian Quiroga, R., Nadasdy, Z., & Ben-Shaul, Y. (2004). Unsupervised spike detection and sorting with wavelets
and superparamagnetic clustering.
Flexibility is an Important Factor
when Designing New MEAs
Anette Trobäck*2, Per Köhler2, Fredrik Ejserholm1, Jens Schouenborg1,
Martin Bengtsson1, Lars Wallman1, Cecilia Eriksson-Linsmeier1
Department of Measurement technology and industrial electrical engineering, Lund University Sweden.
Neuronano research center, Lund University Sweden.
[email protected]
Gliosis, neural interface, immune response, multielectrode array
Since the 70’s when neural interfaces were first described, the use and development of neural probes
has only increased and are now used by many groups as a research tool. In the future they may play
an important role in the diagnosis and therapy of several clinical conditions, such as stroke, trauma or
neurodegenerative disorders, by facilitating motor and pain control. However, most neural devices in
use have several shortcomings. They are relatively large, have a low spatial resolution and they are
not fully biocompatible, thus having little chance of long-term stability, a parameter that is crucial
for clinical use.
Studies have shown that the brain tissue reaction may be the cause of the gradual loss of electrode
recording capacity. This could be a result of the encapsulation tissue surrounding the probe, keeping
the probe insulated from the neurons. In light of these results it is apparent that new materials and
electrode designs are needed to minimize this response and make chronic implantations possible.
We have previously reported developing a new type of micromachined multielectrode array (MEA)
based on a polymer resin (SU-8) [1], designed to minimize the tissue reaction around the bulk of the
implant. Earlier findings show that the tissue reaction around a tethered implant is largest in the
rostro-caudal direction of the CNS, they also show that size affect the gliosis[2].Our MEA is at least
an order of magnitude thinner than commercially available MEA:s, and therefore flexible in the
preferred direction, i.e. in the rostro-caudal direction.
Here we provide results from polymer electrode shanks implanted in the cerebral cortex for 6 and 12
weeks in adults Sprague-Dawley rats. Electrode shanks were implanted either along or perpendicular
to the principal direction of movement of the brain, and either coated in lubricating gelatin or noncoated. Neurons, activated microglia and reactive astrocytes were visualized using NeuN, ED1 and
GFAP antibodies in immunohistochemical analysis. Results were quantified in concentric regions of
interest. Neuron count and GFAP intensity improved when the electrode shanks were implanted
perpendicular to the direction of movement of the tissue, in order to be able to flex with movements.
Gelatin lubrication during implantation produced no significant effect on tissue response. This
indicates that shear stress produced by micro motions in the tissue, but not during the implantation of
the shanks, causes neuron death and astrogliosis.
As today, the method of choice for electrical contact with a neural implant is by anchoring the
electrode to the skull; our flexible approach seems to be the most viable solution for achieving the
smaller rostro-caudal impact thereby achieving a more lasting neural interface.
[1]Kohler, P., et al., Flexible multi electrode brain-machine interface for recording in the
cerebellum. Conf Proc IEEE Eng Med Biol Soc, 2009. 2009: p. 536-8.
[2]Thelin, J., et al., Implant size and fixation mode strongly influence tissue reactions in the CNS.
PLoS One, 2011. 6(1): p. e16267.
An Additive Instantaneously Companding Readout System for
Cochlear Implants
Cees-Jeroen Bes1*, Wouter A. Serdijn1, Jeroen J. Briaire2, Johan H.M. Frijns2
TU Delft, Delft, The Netherlands
LUMC, Leiden, The Netherlands
[email protected]
Cochlear Implants, Additive Instantaneously Companding, Neural Recording, Analog to Digital
Converter, Evoked Compound Action Potential.
Major Cochlear Implant manufacturers have included the possibility of recording neural responses.
However, the possibilities are severely restricted due to the occurrence of saturation in the single
channel amplifier and analog to digital converter (ADC), and the relative high noise levels. This is
most clearly illustrated by the fact that objective neural thresholds are mostly found at the upper end
of the subjective electrical dynamic range (Hughes, Brown, Lopez and Abbas, 1999). Recording on
these relative high levels has as major drawback that different neural waveforms originating from
different fibre populations are combined (Briaire and Frijns, 2005). Potentially the neural response
data, thresholds, but also the spread of excitation and neural recovery functions, could provide
insight in what the optimal stimulation strategy should be, and how to program the current levels of
the implant for individual patients. Especially in very young children this should lead to increased
performance. Researchers are now confronted with the limitations of existing neural response
readout systems needed for reading out the evoked compound action potential (eCAP). These
limitations urge the need for a new neural response readout system having a dynamic range of
126dB, that is small, low noise, power efficient and can handle input signals exceeding the supply
voltage. Existing techniques do not offer solutions to meet the above specifications. An overall
readout system design is proposed containing an additive instantaneous companding (a combination
of compressing and expanding) input system, multiplexer, compensation circuit, amplifier and an
ADC in order to record the eCAPs from the stimulated auditory nerve.
Drug delivery characteristics of PEDOT
Christian Boehler*1,2, Maria Asplund1,2
Freiburg Institute for Advanced Studies (FRIAS), University of Freiburg, Germany
Department of Microsystems Engineering, Lab. for Biomedical Microtechnology, University of Freiburg, Germany
[email protected]
PEDOT, Dexamethasone, controlled release, drug delivery
Conducting polymers like poly (3,4-ethylene dioxythiophene) (PEDOT) show promising qualities
for neural sciences regarding their electrical and biochemical properties so that their usability for
various applications has been intensively investigated in the past years. The main focus thereby was
set to the use as coating material on conventional metallic electrodes to reduce the impedance of
recording electrodes or increase the charge delivery capacity of stimulation electrodes in neural
implants. Another aspect, making conducting polymers superior to pure metallic electrodes is the
ability of incorporating pharmaceutically active substances into the polymer that can subsequently be
released by applying an electrical signal to the polymer. This feature enables the spatially confined
modulation of the immune response against an implanted device by releasing an anti-inflammatory
substance e.g. Dexamethasone (Dex) so that the overall performance of an implanted electrode can
be significantly improved.
In this study we analyze the release characteristics of PEDOT:Dex coated electrodes. We
demonstrate that release is coupled to redox cycling of the electrode and thereby temporal control of
drug release is possible. In addition we show that the commonly used evaluation methods lead to an
overestimation of the drug release.
A drug loaded conducting polymer film was electrochemically realized on Pt-electrodes in a typical
3 electrode setup using a platinum counter electrode and a silver reference electrode. The deposition
was done potentiostatically at a potential of +0.9V in an electrolyte of 0.01M EDOT and 0.01M Dex
as anti-inflammatory drug. Samples of various film thickness could be achieved by controlling the
deposition charge with a potentiostat. For releasing the drug from the polymer, cyclic voltammetry
(CV) in the range of -0.6V to 0.9V was used to switch the polymer between its reduced and oxidized
state in a phosphate buffered saline (PBS) solution. This solution was replaced every 5 scans by new
PBS whereas the ‘old’ solution was used to quantify the amount of the released Dex using the
common UV absorption reading at 244nm but also a high performance liquid chromatography
(HPLC) analysis. Results were compared with an electrochemical quartz crystal microbalance
measurement of equivalent PEDOT:Dex films under the same release conditions.
Results and discussion
All measurement methods confirm that there is a release of the incorporated drug molecule that can
be controlled to some extent, meaning there is a low passive release rate without stimulating the film
compared to an increased amount of molecules that get released during the CV cycling. The release
characteristic thereby shows that the amount of released drugs decreases with increasing number of
CV cycles due to the limited amount of drugs in the polymer (Figure 1). The chromatographic
analysis of the release solution however does not only show a release of the drug but also the
expulsion of EDOT molecules from the polymer film during the CV cycles that cannot be selectively
detected by the other methods. The amount of EDOT thereby is approximately 10 times higher than
that of the simultaneously released Dex. As the EDOT molecules are also UV active in the same
range as the Dex molecules (244nm) this means, that the quantification of the drug in the absorption
reading is significantly affected by the unwanted EDOT release. Based on that, the amount of the
drug that commonly has been determined from absorption measurements in release studies reported
by others [1,2] is considered to be significantly overestimated. The same effect also holds true for the
quantification in the EQCM as this method only detects the overall mass change of the polymer film
without differentiation between various molecules that leave the film but also molecules that might
enter the film to balance the ion movement are not considered in this method. As a result one can say
that the evaluation method for quantifying the amount of released substances from conducting
polymers is crucial due to an influence of the method itself on the results. While EQCM and
absorption measurement lead to an overestimation of the released target molecules, the ‘true’ value
can only be determined by a chromatographic analysis that enables a selective quantification of
different substances in the release solution.
Here we demonstrate the possibility to use electropolymerized PEDOT:Dex as a coating material for
the controlled release of drugs from neural implant microelectrodes. It is shown that timing and mass
of release can be controlled by the number of redox cycles and the thickness of the coating. An
important finding is that significant quantities of the monomer EDOT are eluted simultaneously. This
leads to severe overestimation of released quantities if their analysis is solely based on UV
absorption. More importantly, the unintentional release of EDOT is a major concern considering the
highly reactive monomer could have adverse effects on cells in the vicinity of the implant. Future
work will aim to further study the effect of the combined EDOT:Dex release and investigate
strategies to optimize the control of Dex release and simultaneously prevent EDOT leakage.
Furthermore, the possibility to use this method for release of other drugs of various size and charge
will be investigated.
Figure 4: Drug release characteristic of a PEDOT:Dex film by different measurement methods
1. Abidian, M.R., et al., Conducting-Polymer Nanotubes for Controlled Drug Release. Adv.
Mater.18: 405–409, 2006
2. Xiao, Y., et al., New carbon nanotube-conducting polymer composite electrodes for drug
delivery applications. Polym. Int. 61: 190-196. doi 10.1002/pi.3168, 2011
Advanced recording, stimulation and sensing using PEDOT-CNT MEA
Ramona Gerwig*, Paolo Cesare, Udo Kraushaar, Alfred Stett, Martin Stelzle
Natural and Medical Sciences Institute at the University of Tuebingen, NMI, Reutlingen, Germany
[email protected]
CNT, PEDOT, recording, stimulation, dopamine.
Electrical recording and stimulation of brain tissue in vitro and in vivo [1-3] typically uses
microelectrode arrays (MEA) fabricated from metal based materials [2, 4]. In order to improve signalto-noise ratio, stimulation and sensing properties as well as cell viability, microelectrodes were
modified with poly(3,4-ethylenedioxythiophene) (PEDOT) and carbon nanotubes (CNTs). Previous
reports have demonstrated the suitability of PEDOT as a material for micro-neural interfaces [5].
Furthermore, exceptional viability of cells and their efficient integration with layers composed of
CNTs have been observed [6-8]. The large porosity and effective surface area of CNT electrodes result
in very favorable charge transfer capabilities [6]. Moreover, CNTs modified electrodes provide for
enhanced sensing properties towards neurotransmitters such as dopamine [9]. PEDOT-CNT
composite materials are therefore considered attractive candidates for the fabrication of electrodes in
neuroprostheses as well as for in vitro cell culture systems.
Gold microelectrodes of MEAs were coated by electropolymerization of EDOT in an aqueous
solution also containing poly(sodium-p-styrenesulfonate) and CNT. Coated electrodes show very
low impedance (< 20 kΩ for a 30 µm electrode diameter) and very high capacitance (4-10 mF/cm2)
when compared to state-of-the-art MEA electrodes. They can be produced in a reproducible quality
and withstand intensive rinsing with solvents, autoclaving, UV irradiation, and repeated use in cell
Cardiomyocytes exhibit excellent viability and activity on PEDOT-CNT electrodes over a cultivation
duration of up to 10 days. Recorded signals show increased amplitudes of the QRS-complex and well
defined T-waves (Fig. 1). Recordings from dorsal root ganglia show a reproducible reduction of
noise by 30% as compared to state-of-the-art electrodes which enables the detection of very small
signals (Fig. 1). Cortical networks could be cultivated for at least two months which proofs excellent
biocompatibility of PEDOT-CNT MEAs. Preliminary results show improved signal-to-noise ratio
and a significant increase in signal amplitude of recorded action potentials.
Apart from improved recording, PEDOT-CNT MEAs exhibit significantly high charge transfer
capacitance which provides for the possibility of stimulating cells with extremely low voltages
avoiding tissue damage and side reactions. Constantly high charge per pulse (e.g. 0.8 (PEDOT) and
1.2 mC/cm2 (PEDOT-CNT) with an anodic pulse of 500 µs and 500 mV amplitude) can be
transferred over 3.6 million pulses without impairing the electrochemical characteristics and without
delamination of the coatings.
The improved recording and stimulation electrode can additionally be used as neurotransmitter
sensor. Sub-micromolar concentrations of dopamine and other neurotransmitters could be detected
using square wave voltammetry at PEDOT-CNT electrodes. Selectivity in the presence of ascorbic
acid and uric acid was examined.
In summary, results demonstrate exceptional performance of PEDOT-CNT microelectrodes and their
suitability for applications in in vitro neurotechnology as well as in neuroprostheses [10].
E [µV]
200 400 600
amplitude [mV]
40 TiN
E [µV]
200 400 600
200 400 600
t [ms]
t [s]
Fig. 1: Left: Recordings from autorhythmic activity of primary embryonic chicken cardiomyocytes
on Au, PEDOT and PEDOT-CNT MEAs. Right: Recordings from rat dorsal root ganglia after
stimulation with capsaicin.
S. F. Cogan, Biomedical engineering 2008, 10, 275.
M. D. Johnson, R. K. Franklin, M. D. Gibson, R. B. Brown, D. R. Kipke, Journal of Neuroscience Methods
2008, 174, 62-70.
D. R. Kipke, W. Shain, G. Buzsaki, E. Fetz, J. M. Henderson, J. F. Hetke, G. Schalk, Journal of Neuroscience
2008, 28, 11830.
K. D. Wise, D. J. Anderson, J. F. Hetke, D. R. Kipke, K. Najafi, Proceedings of the IEEE 2004, 92, 76-97.
S. J. Wilks, S. M. Richardson-Burns, J. L. Hendricks, D. C. Martin, K. J. Otto, Frontiers in Neuroengineering
2009, 2.
E. W. Keefer, B. R. Botterman, M. I. Romero, A. F. Rossi, G. W. Gross, Nature Nanotechnology 2008, 3, 434439.
V. Lovat, D. Pantarotto, L. Lagostena, B. Cacciari, M. Grandolfo, M. Righi, G. Spalluto, M. Prato, L. Ballerini,
Nano Letters 2005, 5, 1107-1110.
A. Mazzatenta, M. Giugliano, S. Campidelli, L. Gambazzi, L. Businaro, H. Markram, M. Prato, L. Ballerini,
Journal of Neuroscience 2007, 27, 6931-6936.
P.-Y. Chen, R. Vittal, P.-C. Nien, K.-C. Ho, Biosensors and Bioelectronics 2009, 24, 3504-3509.
R. Gerwig, K. Fuchsberger, B. Schroeppel, G. S. Link, G. Heusel, U. Kraushaar, W. Schuhmann, A. Stett, M.
Stelzle, Frontiers in Neuroengineering 2012, 5, 1-11.
The support of NMI-TT GmbH and Multi Channel Systems is kindly acknowledged. Funding:
BMBF Grant no. 01GQ0834.
Microfabrication of Neural Electrode Arrays by Laser-Processing
*Christian Henle 2, Martin Schuettler 1,2, Juan S. Ordonez 1,
Wolfgang Meier 2, Fabian Kohler 1, Thomas Stieglitz 1,2,3
Laboratory for Biomedical Microtechnology, Deptartment of Microsystems Engineering - IMTEK,
University of Freiburg, Germany 2 Cortec GmbH, Freiburg, Germany 3 Bernstein Center Freiburg, Germany
*email: [email protected]
microfabrication, neural interface, electrode array, laser, silicone rubber,
Today, the vast majority of microelectrode arrays for neural recording and stimulation are fabricated
either by hand crafting or by methods of photolithography-based micromachining. While hand
crafted arrays utilize materials that proved biostable over years and decades (silicone rubber,
Parylene C, polyesterimide, stainless steel and Cobalt- or Platinum-based alloys), the degree of
automation, contact density (electrode dimension d ≥ 0.5 mm) and fabrication precision is limited by
manual skills. Photolithography successfully overcame these limitations (d ≥ 5 µm) and is applied
for highly parallelized fabrication of microelectrode arrays since the late 1960s. However, materials
differ from those traditionally used in implants in terms of layer thickness (some 10 nm to some
10 µm) and composition of substrates (polyimide, liquid crystal polymer, silicon). Although these
devices proved over decades to be invaluable tools for neuroscience, predominantly used in animal
experiments, the approval as (active implantable) medical device is a difficult task which succeeded
so far in only one case: The CE marking of the Argus II visual implant by Second Sight in early 2011
after 20 years of research.
As an alternative fabrication method, we developed laser-based micromachining during the past
years [1], expecting CE marking of one of our electrode arrays in early 2013. The fundamental
process is simple and involves a layer of medical grade silicone rubber (PDMS) spin-coated on a
mechanical carrier to a thickness of some 10 µm. A perforated parylene C foil (15 µm, providing
mechanical strength to the array) is laminated to the PDMS and a second PDMS layer is deposited.
On top, a metal foil (stainless steel, MP35N, Pt or Pt/Ir) of 12.5 µm to 25 µm thickness is laminated
and patterned with a laser, leaving electrode sites, interconnection tracks and contact pads behind.
Subsequently, a covering PDMS layer is deposited. Electrode sites and contact pads are exposed and
the contour of the array is cut using laser ablation of PDMS. Finally, the carrier is discarded and
wires (MP35N or Pt/Ir) are welded to the pads of the array before applying PDMS for electrical
Figure 1:
Picture of laser-fabricated neural electrode array for micro-ECoG recordings
The basic laser-fabrication process can be modified by stacking of multiple metal/silicone layers
allowing higher electrode contact densities [2], meander-shaping of tracks providing stretchability of
the electrode arrays [3], integration of a layer of tough parylene C for improvement of robustness [4],
integration of microfluidic channels for drug delivery as well as three-dimensional shaping to build
e.g. cuff-type electrodes [5]. Figure 1 shows an example of an flat electrode array with single-layer
Pt/Ir metallization, meander-shaped tracks and parylene C reinforcement, suitable for recording
micro-ECoGs (electrocorticograms).
In general, the concept of laser-fabrication of neural electrode arrays permits a high degree of
automation and flexibility (avoiding photomasks) and allows the realization of feature sizes of
d ≥ 50 µm. The overall fabrication time can be as short as one day, the tools required are limited to a
laser, a spin-coater and a microwelder for interconnection of wires. The materials used are
exclusively of medical grade and are all well established since decades by traditional pacemaker-type
implanted electrode fabrication.
Schuettler, M., Stiess, S., King, B., Suaning, G.J.: Fabrication of Implantable Microelectrode
Arrays by Laser-Cutting of Silicone Rubber and Platinum Foil. Journal of Neural Engineering,
No. 2, pp. 121-128, doi:10.1088/1741-2560/2/1/013, (2005).
Suaning, G. J., Schuettler, M., Ordonez, J. S., Lovell, N. H.: Fabrication of Multi-Layer, High
Density Micro-Electrode Arrays for Neural Stimulation and Bio-Signal Recording.
Proceedings of the IEEE Neural Engineering Conference, pp. 5-8, (2007).
Schuettler, M., Pfau, D., Ordonez, J.S., Henle, C., Woias, P., Stieglitz, T.: "Stretchable Tracks
for Laser-Machined Neural Electrode Arrays", Proceedings of the IEEE Engineering in
Medicine and Biology Conference, pp. 1612-1615, (2009).
Henle, C., Hassler, C., Kohler, F., Schuettler, M., Stieglitz, T. "Mechanical Characterization of
Neural Electrodes based on PDMS-Parylene C-PDMS Sandwiched System", Proceedings of
the Annual Conference of IEEE EMBS, pp. 640-643, (2011).
Schuettler, M., Schroeer, S., Ordonez, J.S., Stieglitz, T.: "Laser-Fabrication of Peripheral
Nerve Cuff Electrodes with Integrated Microfluidic Channels", Proceedings of the IEEE
Neural Engineering Conference, pp. 245-248, (2011).
S. Khoshfetrat Pakazad1,*, A. M. Savov1, A. van de Stolpe2, S. Braam3 and R. Dekker1, 2
Delft University of Technology, Delft, The NETHERLANDS
Philips Research, Eindhoven, The NETHERLANDS
Pluriomics BV, Leiden, The NETHERLANDS
[email protected]
Stretchable electronics, MEA, Micro electrode arrays, Mechano-biology, Mechanotransduction.
A platform for the batch fabrication of pneumatically actuated Stretchable Micro-Electrode Arrays
(SMEAs) by using state-of-the-art micro-fabrication techniques and materials is demonstrated. The
proposed fabrication process avoids the problems normally associated with processing of thin film
structures on Polydimethylsiloxane (PDMS), by first fabricating the electrodes and electrical
interconnects on the silicon wafer using fine-pitched stepper lithography, and afterwards transferring
the structures to the elastomer. Stretchability is achieved by a novel spiral design for the
interconnects. Experiments demonstrate the biocompatibility of the fabricated devices for in vitro
cell culturing. Stretchable Micro-Electrode Arrays (SMEAs) are becoming increasingly important in
biomedical research since they enable the study of the mechano-biology of cells cultured on the
SMEAs under electro-mechanical stimulation. Currently the fabrication of SMEAs involves either
processing directly on PDMS, which poses serious fabrication challenges, or the use of stretchable
conducting materials which are not compatible with current state-of-the-art micro-fabrication
techniques. Here we demonstrate the feasibility of a manufacturable process which enables low-cost
mass production of SMEAs for high-throughput clinical and pharmaceutical applications.
It is important that the interconnects and electrodes of the SMEA occupy a minimal surface area and
do not alter the surface topography. Therefore the use of coplanar and non-coplanar wavy and
serpentine structures to achieve stretchability is not possible and a novel spiral design is used for the
interconnects. The complete SMEA consists of pneumatically actuated circular PDMS membrane
with 16 symmetrically arranged electrodes, spiral interconnects and micro-grooves to promote cell
adhesion and alignment. The trajectory of the interconnects is designed to be perpendicular to first
principle strain component in the membrane, and mechanically dimensioned to withstand the second
principle strain. Experiments confirm the absence of cracks in the interconnects for a 3.0 mm
diameter membrane, cyclically inflated to 700 μm height, corresponding to an average radial strain
of 15 %. In order to demonstrate the biocompatibility of the device, stem-cell derived
cardiomyocytes are plated on the PDMS membrane and cultured for two weeks. The cardiomyocytes
showed spontaneous contractile activity and well-defined sarcomeric organization as expected.
(Left) Top view of the chip, (Right) Schematic view of the cell stretching system.
Titanium Nitride as a microelectrode material for auditory nerve
stimulation and sensing purposes.
N. S. Lawand1, P. J. French1, J. J. Briaire2 and J. H. M. Frijns2.
Affiliations: 1Delft University of Technology, Delft, The Netherlands
Leiden University Medical Centre, Leiden, The Netherlands.
Corresponding author: [email protected]
Cochlear Implants (CI’s), Microelectrode array, Titanium Nitride (TiN), Neural stimulation,
COMSOL Multiphysics ® 4.2a.
Cochlear Implants (CI’s) are implantable devices which bypasses the non-functional inner ear and
directly stimulates the hearing nerves with electric currents thus enabling the deaf patients to
experience sound again. Current devices are able to restore hearing far from achieving perfect sound
quality, and so the goal of the research consortium SMAC-it is to develop an improved Cochlear
Implant (CI) with more stimulation sites on the electrode array to address low frequency sound and
with integrated sensors and actuators to reduce the insertion trauma. By photolithography and silicon
microfabrication technology, it should be possible to achieve greater functionality at lower cost.
Microelectrode array, the main component is placed in close contact with neurons to provide reliable
excitation to the auditory nerve. Platinum (Pt), Iridium (Ir) and Platinum-Iridium (Pt-Ir) are the
favorable materials used for neural stimulating electrodes. They show excellent electrical and
mechanical properties suitable for neural stimulation and recording/sensing purposes [1]. But their
long term stability and performance is still a question especially working in an saline environment
where surface electrochemistry plays an important role for effective charge transfer. Titanium nitride
(TiN) is one of the material which exhibits the proficiencies as an microelectrode material with
excellent mechanical, electrical and biocompatible properties ideal for nerve stimulation and sensing
with the capabilities for combination with CMOS process technology and electrical properties. TiN
can be deposited by physical vapor sputter deposition (PVD) techniques with micro-columnar
structure with rough surface which increases the available surface area. Here we present our initial
microelectrode array design, simulation and the fabrication capabilities of sputtered TiN material as a
microelectrode material with respect to its mechanical and electrical properties. Also, we discuss the
initial results for sputtered thin TiN layers and its surface properties suitable for electrical stimulation
and sensing neuronal activity. TiN was chosen because it readily lends itself to reactive sputtering
method and provides significant charge injection rates 23 mC/cm2 [2] with excellent corrosion
resistance and biocompatibility properties.
Simulation, fabrication and characterization of TiN films.
The initial design of the stiff microelectrode array consists of a silicon substrate material with 16 TiN
stimulation sites (Figure 1). The total length is 11 mm with TiN stimulation sites of 75 µm in
diameter with 500 µm pitch. More details regarding design are published elsewhere [3]. Three
different design variations with respect to the stimulation sites are considered as shown in Figure 1.
Stimulation sites with protruded design, planar design and embedded design were studied with the
help of COMSOL Multiphysics ® 4.2a to observe the electric charge density distribution at the
stimulation sites when stimulated under normal conditions as well as saline environment. This field
should have sufficient value to trigger an action potential in the nerve, but should also be low to
avoid neural damage. Electric field distribution at the surface and the potential in the form of contour
lines for the protruded shape design is as shown in figure 2. TiN films were sputter deposited on ptype Si(100) wafers by DC reactive magnetron sputtering from a titanium (Ti) target of 99.999%
purity. At a chamber pressure of 2.106 x 10-8 mbar pure Argon (100 sccm) and Nitrogen (300 sccm)
gases were purged. The working pressure during deposition is between 0.0066 mbar to 0.0133 mbar.
Pre-sputtering of Ti at 5 kW for 2 minutes in argon atmosphere cleans the target surface. Substrate
temperature between 27 and 4000C has insignificant effect on the mechanical properties so the
sputtering was carried at 3000C. To study the RF power effects on crystallography, stress and surface
roughness we deposited TiN at 0.5, 2.5 and 5 kW with different sputtering time to achieve 200 nm
thickness. The substrate bias voltage was kept to 0 V for optimum sputtering conditions.
Variation of TiN mechanical properties are correlated with microstructural characteristics such as
composition, grain size and orientation and density/voids content. To study the microstructural
surface difference, Titanium (Ti) and TiN of 200 nm thick were deposited with above sputtering
conditions. TiN surfaces show porous structure consisting of tapered crystallites with less hilliocks
providing high surface area with increased grain boundaries which favours nerve stimulation since
more available fractal surface area provides high charge transfer which is not possible with smooth
surfaces. Stress in TiN layers was measured using wafer curvature. TiN layers exhibit low tensile
stress for low power and compressive stress for layers deposited at high power. The sheet resistance
was 73.45 Ω/□ which is well within the requirements and further process parameter changes did not
cause deviation from the required electrical properties.
[1] S. F. Cogan, “Neural Stimulation and recording electrodes”, Annual Review Biomed. Eng. Vol.
10, (2008), 275-309.
[2] M. Janders, U. Egert, “Novel thin-film titanium nitride microelectrodes with excellent charge
transfer capability for cell stimulation and sensing applications”, Proc. 19th Int. Conf. IEEE Eng.
Med. Biol. Soc., (1996) 1191-1193.
[3] N. S. Lawand, P. J. French, J. J. Briaire, J. H. M. Frijns, “Development of probes for cochlear
implants”, IEEE Sensors, (2011), pp. 1827-1830.
Figure 1. Electrode design with three different design configuration for the TiN stimulation site.
Figure 2. Electric field distribution of the model for the protruded shape design.
Elastomer Based Neural Electrodes for Mechanically Challenging
Ivan R. Minev, Pavel Musienko, Nikolaus Wenger, Grégoire Courtine and Stéphanie P. Lacour
Centre for Neuroprosthetics, École Polytechnique Fédérale de Lausanne (EPFL), Switzerland
[email protected]
Soft neural electrode, PDMS
Neural interfacing necessitates the use of arrays of electrodes implanted in soft tissue. The materials
and technologies used in the fabrication of neural electrode arrays have been borrowed from the
semi-conductor industry. Often the substrate that carries the electrode array is made from silicon or
flexible (but not stretchable) plastics such as polyimide or parylene. Traditional electrode materials
are Platinum, Titanium or Gold. Although the use of these materials enables the miniaturisation of
dense electrode arrays, the resultant implant remains rigid. Neural tissue however is soft, with elastic
modulus several orders of magnitude smaller, even than that of flexible plastics. The mechanical
mismatch may be the cause of a number of undesired effects associated with chronic neural implants.
Rigid implants cannot deform to accommodate the natural movements of joints and spinal cord
segments or conform to skin stretch, they cannot mould around the curvilinear surface of brain,
spinal cord or the inner surface of the retina. Rigid neural electrodes can cause sheer stress and
irritation at the tissue-implant interface and contribute to a sustained foreign body reaction. These
challenges motivate the incorporation of mechanically soft, biocompatible materials such as silicone
elastomers (e.g. polydimethylsiloxane, PDMS) or hydrogels for implants that remain electrically
active under large deformation. The development of soft neural interfaces would necessitate the
adaptation or development of suitable microfabrication technologies and their successful integration
with similarly compliant electrodes.
We will present a simple dry fabrication process that can be easily tuned for producing devices
intended for various implantation sites where the mechanical compliance of the implant is important.
We demonstrate electrical functionality at large tensile strains1 (during lab bench tests) as well as
recordings with assembled devices in rats. For high signal-to-noise recording applications we have
developed a compliant micro-cuff electrode (Fig 1(a)) monitoring cutaneous, proprioceptive2 and
bladder afferent3 signals in the spinal cord.
An important application where a soft and stretchable electrode is of paramount importance is
epidural spinal cord stimulation and/or recording. In this case the electrodes reside in the space
between the spinal canal (elastic modulus of 10s of GPa) and the spinal cord (elastic modulus of 0.1 10kPa). To avoid compression of the spinal cord during animal movement caused by a rigid or
merely flexible implant, it is necessary to utilize thin films that can conform to non-zero Gaussian
curvatures. This motivates the need for a stretchable electrode array which we achieve through the
use of PDMS elastomer and sub micrometer thick Gold electrodes. We demonstrate the results of
preliminary chronic epidural stimulation experiments in rats implanted with the device illustrated in
Fig 1(b).
200 μm
Fig 1 PDMS based neural interfaces.
(a) micro-cuff interface used to record afferent neural activity in dorsal roots. A nerve strand is surgically
inserted in-vivo in the microfluidic channel. (b) planar electrode array wrapped around a cylindrical plastic rod.
The rod is bent to simulate movement of the spinal cord of a rat causing no tearing or creasing of the
electrode array. Square size is 4mm.
I. R. Minev, S. P. Lacour, Applied Physics Letters 2010, 97, 043707.
I. R. Minev, D. J. Chew, E. Delivopoulos, J. W. Fawcett, S. P. Lacour, Journal of Neural Engineering 2012, 9,
E. Delivopoulos, D. J. Chew, I. R. Minev, J. W. Fawcett, S. P. Lacour, Lab on a Chip 2012, 12, 2540.
Advances in polyimide-based thin-film electrode arrays
Juan Ordonez1,2*, Tim Boretius1, Eva Fiedler1,2, Martin Schuettler1,2, Thomas Stieglitz1,2
Affiliations: 1. University of Freiburg, Laboratory for Biomedical Microtechnology, Freiburg, Germany
2. Bernstein Center Freiburg, Freiburg, Germany
[email protected]
Thin-film, polyimide, electrode, adhesion, neuroprosthetic
The use of micromachining technologies for fabricating neural interfaces has led in the last decades
to novel silicon and polymer based (micro-) implants. Polyimide was an early candidate for flexible
implants and a great deal of work was invested to study different designs, material stability and
various aspects of biocompatibility as well as the functional outcome of the different electrode
designs in vivo. For example, sieve electrodes to interface regenerating nerves, multichannel cuff
electrodes but also newer concepts like longitudinal and transverse intrafascicular multichannel
electrodes (LIFE, TIME) have been developed with polyimide as substrate and insulation material.
For interfacing the central nervous system (CNS), planar electrode arrays have been developed to
record local field potentials in rodents and monkeys. Evaluation in vitro as well as in vivo of the
material itself but also of the different electrode designs and concepts showed good material stability,
no material toxicity, only little foreign body reaction after implantation and good spatial selectivity
of the electrodes depending on the chosen concept to interface the PNS and CNS in chronic
Adhesion failure of the polymer-metal laminates has hindered the implementation of thin-film based
electrode arrays in long-term applications in vivo. Delamination frequently occurs at the opened
interfaces of metal and polymer, commonly used as active sites or interconnections. Silicon carbide
(SiC) was introduced as adhesion promoting layer between the polymer and metal. The layers were
deposited through plasma-enhanced chemical vapor deposition (PECVD), allowing a transitive
deposition of polymeric carbon (sp2 hybridization) into ceramic carbon (sp3 hybridization). The
metallic tracks were individually coated with a conformal cladding. The 50 nm thick coating
provides a chemical transition from the polymer into the metal allowing a chemical adhesion at every
present interface. Using such thin layers does not compromise the flexibility of the arrays, but
increases the mechanical stability by allowing any external forces to be dissipated into the stable
polymer substrate, protecting the delicate metallic wiring.
The adhesion was proven mechanically and the interfaces were analyzed by means of x-ray
photoemission spectroscopy, providing enough evidence for the existence of chemical bonds. The
stability the fabricated electrode arrays were controlled in a long term pulse test study in vitro, in
which the devices remained unaffected after two billion stimulation pulses. (I = 375 uA, t = 200 us).
In this work, we will outline the latest approaches to manufacture a long-term stable polyimide-based
neural interface with high charge-injection coatings (e.g. sputtered iridium oxide films) to be used in
Figure 5: Focused ion beam cross-section image of a metallic track. The SiC coating conformally coats the metallic wires and
provides chemical adhesion between the materials at all interfaces.
Sputter Deposited TiN Thin Film Coatings for Optimized
Electrochemical Performance of Neurostimulation Electrodes
S. Sørensen1*, I.H. Andersen1, K. Rechendorff1, L.P. Nielsen1, S. Meijs2, M. Fjorback2
Danish Technological Institute, Aarhus, Denmark
Neurodan A/S, Aalborg, Denmark
[email protected]
Magnetron Sputtering, Titanium Nitride Coating, Neurostimulation, Electrode.
Titanium nitride (TiN) coatings have traditionally been used to improve the tribological performance
of cutting tools. However, the excellent mechanical strength, high electrical conductivity and
biocompatibility of TiN coatings also make them attractive for neurostimulation electrodes intended
for chronic implantation in the human body. Porous TiN coatings, with high surface-to-area ratios,
are already being used extensively for cardiac pacing electrodes where both safety and efficacy are
well documented [1]. Furthermore, a number of plasma assisted surface modification and coating
techniques are available to produce, in principle, the desired surface topography [2].
The aim of the present study is to develop porous TiN coatings with state-of-the-art performance for
a neurostimulation application. The TiN coatings are synthesized by reactive DC magnetron
sputtering (Physical Vapor Deposition) using industrial scale deposition equipment.
Sputter deposition enables tailoring of the coating properties, e.g. stoichiometry and morphology, by
varying a number of different parameters such as sputtering power, deposition temperature,
deposition pressure, coating thickness and substrate bias voltage. The latter controls the kinetic
energy of the ion flux bombarding the growing film. The thin film coatings have been characterised
by X-ray diffraction (XRD), scanning electron microscopy (SEM), energy dispersive x-ray
spectroscopy (EDX) and Rutherford backscattering spectroscopy (RBS). Furthermore, the
performance of TiN coated electrodes has been investigated electrochemically by electrochemical
impedance spectroscopy (EIS), voltage transient measurements (VTM) and cyclic voltammetry
A series of electrodes (surface area: 6 mm2) were coated with TiN by the method described above,
using different deposition parameters. These coatings had an impedance magnitude ranging from 644
Ω to 1.8 kΩ at 100 mHz, as compared to > 100 kΩ for smooth TiN. CV results showed that the overstoichiometric TiN electrodes have a cathodic charge storage capacity (CCSC) which is 89 to 295
times larger than the CCSC of a smooth TiN electrode. Charge injection limits could not be
established for any of the porous TiN electrodes. The charge injection limit of a stoichiometric TiN
electrode was 4 mA. CV and EIS showed consistent electrochemical performance for the different
electrodes in terms of impedance magnitude and CCSC. The performance of the same electrodes in
VTM, in terms of cathodic voltage excursions at 20 mA, was different. At last a single electrode
series was fabricated using two consecutive depositions. This electrode performed best in all
electrochemical tests. It had an impedance magnitude of 325 Ω at 0.1 Hz, a CCSC 625 larger than
the CCSC of a smooth TiN electrode and its voltage excursion at 20 mA was also smallest of all
over-stoichiometric electrodes.
Figure 6: Scanning electron microscopy (SEM) image of a porous titanium nitride coating. The insert is a top-view image of
the thin film coatings showing the pyramidal surface morphology.
Preliminary results have shown that optimized porous TiN coated electrodes with surface domed
pyramids of about 500 nm in size (see figure) resulted in substantially lower impedance values as
compared to reference electrodes coated with a smooth TiN in vitro. Results from in vivo studies
using the optimized porous TiN coated electrodes are presented on [3].
This project is funded by the Danish Advanced Technology Foundation.
1. Schaldach M, Hubmann M, Weikl A, et al. Sputter-deposited TiN electrode coatings for superior sensing and pacing
performance. PACE. 1990; 13: 1891-1895.
2. D.M. Mattox, V.H. Mattox (Eds.) 50 Years of Vacuum Coating Technology and the growth of the Society of Vacuum Coaters,
Society of Vacuum Coaters, 2007, Ch.6.
3. S. Meijs, et al., 4th int. conf. on Neuroprosthetic devices, Freiburg, 2012 [Submitted].
A 10MHz Switched-mode power efficient neural stimulator
M.N. van Dongen*, W.A. Serdijn
Delft University of Technology
Delft, the Netherlands
[email protected]
Keywords (up to 5)
Neural Stimulation, Electrode-Tissue interface, Tissue Modeling, Switched-mode operation
Key requirements of implantable neural stimulators include size and reliability. The size can be
reduced by making the system power efficient (i.e. smaller battery) and by reducing the number of
external components. The latter option simultaneously increases the reliability.
This work presents a stimulator design that meets both goals by employing high frequency (10MHz)
switched-mode stimulation. It combines two common system blocks, the DC-DC converter and the
output stage, into a single system block. It therefore requires only one external inductor and the
switched-mode operation ensures a high power efficiency: 100% theoretical efficiency and a 60%
post-layout simulated efficiency over the full output range.
The circuit is implemented in AMS 0.18μ High Voltage Technology. The circuit design is based
around a boost-converter and furthermore includes a duty cycle generator and some basic logic
controlling the two independent output channels. The 3.5V input voltage is compatible with medical
batteries and the output voltage can be as high as 10V.
Furthermore the current source output characteristic of the system makes charge cancellation easily
implementable for safety reasons. Also the high frequency output signal was shown to be capable of
achieving effective stimulation: a discrete component realization of the system was successfully
tested in practice.
A Study on Modeling of Iridium Electrodes
Naser Pour Aryan*, 1Hans Kaim, 2Sebastian Schleehauf, and 1Albrecht Rothermel
University of Ulm, Ulm, Germany
Retina Implant AG, Reutlingen, 72770, Germany
*[email protected]
Iridium electrode, impedance spectroscopy, electrode modeling
Electrode models are important in designing neural stimulator devices as they are necessary for
appropriately optimizing the characteristics of the stimulator circuit. For applications requiring high
charge injection capacities, iridium is a good option as electrode material [1].
In our applications we have used microelectrode arrays (MEAs). The MEAs are manufactured using
thin film lithography on a float glass substrate. Gold lines are patterned with a lift-off technique on
the surface of the substrate. The metal lines are covered with a polyimide insulation layer. A hard
mask is used to remove the polyimide at the electrode sites and over the conduction pads. The
iridium electrodes are patterned in a subsequent lift-off process. Every MEA contains 59 circular
iridium electrodes having a diameter of 15µm.
The solution used in the experiments was phosphate buffered saline (PBS). We used the potentiostat
“VersaSTAT 4” for impedance spectroscopy to extract the impedance magnitude and phase versus
frequency for every electrode. The voltage sine wave amplitude was 10mV. The chosen frequency
range was 100 Hz to 1 MHz to achieve a consistent model. For higher or lower frequencies other
models would be necessary. We then found models fitting the extracted frequency response using the
software ZView 2 (© 2005 Scribner Associates, Inc.). Fig. 1 shows the electrode model
approximated for the electrodes. This model does not include the Warburg (diffusion) impedance
which is due to diffusion of chemical reactants in the solution. The Warburg impedance is only
considerable for lower frequencies [3]. The model has also been introduced in [2], but there, other
materials, i.e. platinum, platinum black and titanium nitride, were investigated.
The constant phase angle impedance ZCPA represents the interface capacitance impedance. RF is the
faradaic resistance representing the current flowing into the electrode due to chemical reactions. RS is
the solution spreading resistance.
ZCPA is a measure of the non-faradaic impedance arising from the interface capacitance and is given
by the empirical relation [3]:
T is a measure of the magnitude of ZCPA. P is a constant (
) representing inhomogeneities in
the surface and
. When P=1, ZCPA represents a pure capacitor corresponding to the interface
capacitance. The parameters P and T depend on the electrode material [4].
Figure 7: The extracted electrode model
Table 1 lists the mean extracted electrode parameter values and the corresponding standard deviation
percentages. 45 electrodes were measured from one MEA.
Table 1: Model parameter values
Mean value
Standard deviation normalized to mean value
31.5 %
3.57 nF
26 %
2.8 %
8.33 MΩ
60.6 %
[1] N. Pour Aryan, C. Brendler, V. Rieger, S. Schleehauf, G. Heusel, and A. Rothermel. „In vitro
study of iridium electrodes for neural stimulation”. In Engineering in Medicine and Biology
Society, EMBC, 2012 Annual International Conference of the IEEE, 28 2012-sept. 1 2012.
[2] Franks, W.; Schenker, I.; Schmutz, P.; Hierlemann, A.; , "Impedance characterization and
modeling of electrodes for biomedical applications," Biomedical Engineering, IEEE
Transactions on , vol.52, no.7, pp.1295-1302, July 2005
[3] E.T. McAdams, A. Lackermeier, J.A. McLaughlin, D. Macken, J. Jossinet, “The linear and
non-linear electrical properties of the electrode-electrolyte interface,” Biosensors and
Bioelectronics, Volume 10, Issues 1
[4] Poppendieck, Wigand., "Untersuchungen zum Einsatz neuer Elektrodenmaterialien und deren
Evaluation als Reiz- und Ableitelektrode." Universität des Saarlandes : Dissertation
Naturwissenschaftlich-Technischen Fakultät II - Physik und Mechatronik -, 2009.
Neuroblastoma Cell Proliferation on Hydrogen and Oxygen
Terminated Nano-crystalline Diamond
Matthew McDonald*, Aida Vaitkuvienè, Vilma Ratautaite, Farnoosh Vahidpour, Ken Haenen, Milos
Nesladek .
Affiliations: IMO-MEC, Diepenbeek, Belgium
[email protected]
Nano-crystalline diamond, neuroblastoma, biocompatibility, cell viability, cell adhesion.
Development of highly functional interfaces between biological environments and semiconductor
devices is a crucial area of research in order to create advanced cell-based biosensors and neural
microelectrode array devices. Diamond has been praised as an ideal material for bioelectronics due
to its biological, but also, electrical and mechanical properties1. Based on its unique chemical
properties, the surface of diamond can be functionalized in countless ways using carbon chemistry
routes2. However, the reaction of diamond surfaces to biological environments can be highly
specific, depending on the diamond surface termination as well as on the cell type. To use diamond’s
excellent properties, the device-cell interface have to be optimized first. To do so we investigate the
biocompatibility of the two most common functionalizations: hydrogen and oxygen terminated, in
both intrinsic and boron doped nano-crystalline diamond.
Nanocrystalline diamond was grown in hydrogen plasma with a 1% methane concentration at 700C
using an Astex PE CVD system. The diamond layers were grown approximately 150nm thick on
planar quartz substrates as well as on gold substrates with protruding microstructures.
Neuroblastoma cells were cultivated on the substrates and cell-adhesion and proliferation studies
were performed. Trypan blue exclusion test was conducted for cell-adhesion studies and MTT assay
was performed for viability test. The results showed that although there is a major difference
between contact angle of hydrogen and oxygen terminated diamond at the cell experiment start, the
contact angles quickly converge due to adsorption of proteins on the surface, leading to similar
results on both adhesion and viability tests. Additionally, there appears to be no difference in cell
proliferation between intrinsic NCD and boron doped NCD, nor for planar or protruding surfaces.
All samples showed good biocompatibility with cells lasting for at least 3 days and is congruent to a
similar study done using Chinese hamster ovarian (CHO) cells3. From the results obtained, it seems
that the protein absorption from the medium in the early stage of the cell is detrimental. To trace the
mechanism of the protein absorption, electrospray analysis has been carried out after washing the
cells from variously terminated substrates after the cell cultivation. Both the cultivation from serum
free and serum containing media have been compared, showing enhanced absorption of specific
proteins expressed at the surface that are responsible for the contact angle changes.
[1] Kotov, Nicholas a. et al. “Nanomaterials for Neural Interfaces.” Advanced Materials 21.40
(2009): 3970–4004. Web. 10 Aug. 2012.
[2] Rezek, Bohuslav et al. “Diamond as Functional Material for Bioelectronics and Biotechnology.”
(2011): n. pag. Print.
[3]Smisdom, Nick et al. “Chinese Hamster Ovary Cell Viability on Hydrogen and Oxygen
Terminated Nano- and Microcrystalline Diamond Surfaces.” Physica Status Solidi (a) 206.9
(2009): 2042–2047.
Investigation of electrochemical behavior of
porous TiN stimulation electrodes in vivo
S. Meijs* 1, M. Fjorback2, S. Sørensen3,
K. Rechendorff3 and N.J.M. Rijkhoff1
Aalborg University, Aalborg, Denmark; 2Neurodan A/S, Aalborg, Denmark; 3Danish Technological Institute, Århus,
[email protected]
Stimulation electrodes, electrochemical characterization, voltage transient measurement.
It is important to avoid stimulation pulses which are not within safe charge injection limits. For most
materials the charge injection limits are the water reduction or oxidation potentials.1 However, these
limits are electrolyte specific, meaning that limits established in vitro may not apply to the in vivo
situation. Although it is possible to establish such limits, in practice they are not always obeyed.2
In this study, voltage transient measurements (VTM) were made in vivo and in vitro before, during
and after a 90 days implantation period. Differences in the voltage transients and charge injection
limits are reported.
Electrode fabrication
15 titanium electrodes (surface area: 6 mm2) and 3 titanium discs (surface area: 300 mm2) were
coated with titanium nitride (TiN) using reactive pulsed magnetron sputtering to create an overstoichiometric coating with a thickness of 12.5 µm. The electrodes were produced using optimized
production settings to obtain a coating with the highest possible charge injection capacity without
compromising mechanical performance. These settings were established in a previous study.3
In vivo and in vitro and passive and active electrodes were used to distinguish between the effect of
implantation and active use. 10 electrodes were used in vivo, 4 of which were used actively, while
the remaining 6 served as passive implants. 5 electrodes were used in vitro, 4 of which served as
active electrodes and 1 as a passive electrode.
Experimental protocol
10 electrodes and 2 counter electrodes were implanted in 2 Göttingen minipigs. 4 electrodes had
percutaneous connections to allow for in vivo measurements. In vitro electrodes were kept in ¼strength Ringer solution at 37.5 °C. The in vivo measurements consisted of electrochemical
impedance spectroscopy (EIS), electrical stimulation and VTM. The same protocol was used in vivo
and in vitro to ensure that the only difference was due to the electrolyte.
Before and after implantation the electrodes were characterized using electrochemical and analytical
methods. Electrochemical characterization was done using VTM, EIS) and cyclic voltammetry (CV)
in ¼-strength Ringer solution at room temperature. EIS was performed using Solartron, Model 1294
in conjunction with 1260 Impedance/gain-phase Analyzer (Solartron Analytical, UK). CV and VTM
were performed using VersaSTAT 3 potentio- galvanostat (Princeton applied research, USA).
Voltage transient measurements
VTM were made using a cathodic-first bipolar symmetric stimulation pulse; each with a phase width
of 200 µs. Charge injection limits were determined using the maximal cathodic and anodic values
from which the potential drop due to the electrolyte resistance (IR-drop) was subtracted. The IR-drop
was calculated using the potential at 20 µs after pulse cessation and subtracting it from the last data
point of the anodic phase. The charge injection limit was reached at -0.6 V or 0.8 V vs. Ag|AgCl. As
no reference electrode was implanted, the in vivo charge injection limits were -0.403 V and 1.007 V.
Pre-implantation measurements
Charge injection limits could not be measured in vitro due to the low slew rate of the VersaSTAT 3
at currents exceeding 20 mA. The cathodic peak potentials of the voltage transients of the in vitro
electrodes, however, were significantly lower than the cathodic peak potentials of the in vivo
electrodes before implantation. This indicates a better performance of the in vitro electrodes. The
same difference was observed in EIS and CV. The in vitro and in vivo electrodes had a mean
electrochemical surface area of 26 and 18 cm2, respectively.
In vivo measurements
VTM of the in vivo electrodes showed
increasing voltage excursions over time, as
shown in figure 1. However, the increasing
trend ceased after day 47. This may indicate
that the electrode-tissue interface had
stabilized, although the amount of data is not
sufficient to draw firm conclusions.
Furthermore, the voltage excursions measured
in vivo were at the beginning of the study 14
times larger and at the end 26 times larger
than the voltage excursion measured in vitro.
The in vivo charge injection limits,
determined at the termination of the in vivo
study, ranged from 0.2-1.0 µC. Charge
injection limits could not be established for
the in vitro electrodes as pulse amplitudes
exceeded 20 mA.
Figure 1: Voltage transients of a single electrode measured in
vivo at several instances after implantation using a biphasic
current pulse with amplitude 5 mA. The IR drop has been
subtracted and the open-circuit potential was set to 0 V for
Post-implantation measurements
After 90 days, there was no longer a significant difference between in vivo and in vitro electrodes
and neither was a difference in performance found between the active and passive electrodes.
The difference between in vivo and in vitro electrodes before implantation was due to contamination
with silicone adhesive, as the in vivo electrodes had a silicone jacket with fixation mechanisms. After
explantation, the jacket was removed and the difference between the electrodes diminished.
The voltage excursion differed in vivo and in vitro, which is in accordance with results from Wei and
Grill.2 Novel in this abstract is that VTM were made in vivo up to 3 months after implantation and it
was found that the voltage excursions increased over time. We hypothesize that this is due to
adhesion of fibroblasts on the TiN surface as a capsule is formed around the implant.4 The potential
excursions appeared to stabilize from day 46 on, which could signify the fact that a stable capsule
has been formed.
This project is funded by the Danish Advanced Technology Foundation.
1. S. Cogan, Annu. Rev. Biomed. Eng. Vol. 10, 2008, pp. 275-309.
2. X. Wei, W. Grill, J. Neural Eng, Vol. 6, 2009, pp. 6-9.
3. S. Sørensen, et al., 4th int. conf. on Neuroprosthetic devices, Freiburg, 2012 [Submitted].
4. B. Groessner-Schreiber, et al., J. Biomed. Mater. Res. Vol. 64, 2003, pp. 591-599.
Selective Stimulation of the Vagal Nerve Reduces Blood Pressure and
Avoids Significant Bradycardia and Bradypnea
Dennis T.T. Plachta*1, Oscar Cota1, Nayeli Espinosa1, Thomas Stieglitz1,3, Mortimer Gierthmuehlen2
Laboratory for Biomedical Microtechnology, Department of Microsystems Engineering-IMTEK, University of Freiburg,
Freiburg, Germany
Department of Neurosurgery, University Medical Center Freiburg, Freiburg, Germany
Bernstein Center Freiburg, University of Freiburg, Freiburg, Germany
Contact: [email protected]
BaroLoop, blood pressure reducing implant, baroreflex, vagus nerve stimulation (VNS),
Hypertension, a typical disease of civilization, a worldwide threat to patient’s health and a burden to
health care systems. While most patients can rely on pharmaceutical treatments, this option is
ineffective for some due to compliance or ineffectiveness of drugs. Electrical stimulation of afferent
nerve fibers originating from pressure sensors can trigger the baroreflex to reduce blood pressure and
might be an alternative to treat patients with refractory hypertension. Stimulations targeting the
vagus nerve (VNS) are common practice for severe epilepsy therapy and are considered a safe
implantation technology. Downside is, the vagal nerve is the major communication path for many
organs. Robust bipolar stimulation of the complete cross section of the vagal nerve results in
secondary effects such as chest and throat pain, cough, voice alteration, nausea, vomiting and even
arrhythmia. The localization of blood pressure relevant fibers is therefore the first step to embark on
selective stimulation of such fibers. For our future BaroLoop implant, these tasks have to be
accomplished by a fully implantable device, whose available computing power is limited using
embedded systems such as MSP430 (Texas Instrument) microcontrollers.
In an experimental series we used three male Wistar rats. The initial anesthesia was established with
2-4 % Isoflurane combined with subcutaneous application of Carprofen (5 mg/Kg bodyweight). The
ongoing anesthesia was maintained by 1–2% Isoflurane (controlled by the respiration rate). During
experiment the rat was placed on a controlled heat pillow. On an hourly base the rat was applied
isotonic Ringersolution (1ml/100g bodyweight, s.c.). The left vagus and the attached carotid artery
were exposed using a medial cut along the throat. The carotid artery was distally ligated, while a tip
catheter (JCP Codman, 3F) was introduced proximally, pushed to the aortic arch and attached. In the
next step our Cuff-electrode [1] was warped around the vagus without any further alignment. A onechannel ECG recording was established using two subcutaneous needle electrodes at the thorax and a
ground electrode at the right foot. All experiments were approved by and conducted in full
accordance with the animal care regulations of the Regierungspräsidium Baden-Württemberg.
Using tripolar recording and coherent averaging we were able to isolate and localize baroreceptive
fibers in the vagus nerve [2]. After localization we used the proximal electrode for selective
stimulation using a charge-balanced signal. The stimulation was tripolar using the center electrode as
the cathode. Best results were found using 30 Hz repetition rate with currents below 1.5 mA.
We were able to trigger the baroreflex and reduce the blood pressure within the given limit of less
than 30 % while simultaneously avoiding significant side effect in terms of bradycardia and
Figure. 1: A and B: Upper chart shows blood pressure (systolic = blue, MAP = red and diastolic = green), middle chart
shows heart rate and lower chart shows stimulus track. In A: A non blood pressure correlated electrode was chosen for
selective stimulation. Note that the blood pressure hardly decreases but respiration and heart rate show significant drops.
In B: A correlated electrode was chosen with similar stimulation parameters. Note that the blood pressure shows a drop
within the wanted limits and side effects in terms of bradycardia and bradypnea show up only in insignificant form if at
In this work, we present our newest results on selective stimulation of blood pressure relevant fibers
of the Vagal nerve in a rat. Extracting the signal with the cuff-electrode presented the challenge of
separating the desired signal from the rest of the signals travelling inside the vagus nerve. For that
purpose, the coherent real-time average resulted in a valuable tool, which eliminated signals that are
not correlated to the heart activity. Once the proximate electrode is identified it can be used for
selective stimulation of the baroreflex. The selectivity can be further increased using all features of
both, the electrode and the stimulation circuit, i.e. pentapolar stimulation and steering currents.
[1] T. Stieglitz, T. Boretius, J. Ordonez, C. Hassler, C. Henle, W. Meier, D.T.T. Plachta, M. Schuettler: Miniaturized
neural interfaces and implants. Proc. SPIE 8251, Jan 21-26, 2012, San Francisco, USA, 82510A, 2012.
[2] D.T.T. Plachta, N. Espinosa, M, Gierthmuehlen, O, Cota, T, C. Herrera, T. Stieglitz: Detection of baroreceptive
activity in rat vagal nerve recording using a multi-channel cuff-electrode and real-time coherent averaging. Anual
Conference of the IEEE EMBS, San Diego, 2012, in press.
Flexible brain implants dressed up with a real biological surface
Anja Richter*1, Charli Kruse3, Sandra Danner3, Andreas Moser1, Ulrich G. Hofmann2
Affiliations: 1 - Department of Neurology, Group of Neurobiochemistry, UKSH, Lübeck, Germany
2 – Neuroelectronic Systems, Department for Neurosurgery, Albert-Ludwigs-University Freiburg, Germany
3 – Fraunhofer Research Institution for Marine Biotechnology, Lübeck, Germany
[email protected]
Keywords (up to 5)
Brain implantation, flexible probes, adult stem cells, immunohistochemical analysis,
Optimized biocompatibility of brain implantable devices is of major interest to improve long-term
connectivity and stability. To minimize the foreign body reaction resulting in an inactivating glial
scar, biological, chemical as well as mechanical parameters of brain tissue and the implant have to be
taken into account. Here, we present results of an approach that addresses the physical mismatch of
skull-tethered, rigid probes with the soft brain tissue on the one hand by using flexible polyimidebased microprobes. On the other hand, we evaluated an additional biologization of the surface using
fibrin as a native surface coating, known for beneficial effects on neuroregenerative processes. As a
third moiety we coated the flexible implant surface with glandular stem cells and covered them with
the fibrin-based hydrogel, as such protected against abrasion on implantation. These cells have
already demonstrated advantageous effects on skin wound healing and showed promising
neuroprotective secretion patterns in vitro. Using them, we tried to create a syngene implant surface
with real biologically responding and integrating features.
By immunohistochemical analysis of the animal experiments in rats over 6 months, we were able to
demonstrate, that the flexible implants themselves elicit a clearly decreased gliosis as compared to
published reactions on rigid probes. The biologization of the implants using adult glandular stem
cells furthermore cut the required stabilization-time of the tissue regeneration by half to 12 weeks. At
that point no critical encapsulation by astrocytes was visible and the neuronal tissue was found in
close proximity to the implant. It is worth mentioning, that all animals tolerated the implant without
complications or inflammation until the planned termination.
Consequently we conclude that camouflaging the flexible probe with the body’s own cells further
improves the brain tissue response to the prosthetic device and accelerated healing.
Wireless Multichannel Myoelectric Implant for Control of Prostheses
Daniel McDonnall*, Christopher Smith, Scott Hiatt,
Ronald Madsen, Andrew Wilder, K. Shane Guillory
Ripple, Salt Lake City, USA
[email protected]
Myoelectric, electromyography, prosthesis control, wireless, recording implant
Control of prosthetic arms has been limited by the small number of input signals that are currently
used to control multiple degrees of freedom in the limb. Conventional myoprostheses rely on a pair
of signals recorded from surface electrodes on the residual limb. We are developing an implantable
multichannel myoelectric device to detect signals from multiple residual muscles that will be sent
wirelessly to the prosthetic limb. This approach offers the advantages of recording more channels of
independent muscle signals and providing access to deep muscles that cannot be detected with
surface electrodes. We report the results of a proof-of-concept study to verify the in vitro
performance of the system and an in vivo trial to validate device function in an animal model.
The implant was constructed on a ceramic composite circuit board with a bioamplifier ASIC and
additional discrete components. The implant was inductively powered by an external transceiver,
and digitized signal data were sent from the implant by reflected impedance modulation. Each
implant included four pairs of electrodes in epimysial disc, intramuscular bands, and fine wire
configurations. The electronic components and ASIC die were coated with a conformal electronics
sealant, and the entire assembly was coated in silicone.
Benchtop implant performance was verified both in a dry configuration and while the devices were
soaked in saline. The amplifier was shown to have an input-referred noise of 2.2+0.03 μVRMS, a
common mode rejection ratio greater than 55 dB, and neighboring channel isolation averaging 66+8
These prototype implants were validated in a six-dog study at the University of Utah. Two fourchannel devices were implanted bilaterally in the front limb by placing the electronics package
behind the shoulder blades with electrodes implanted in deltoideous and lateral head of triceps. One
week following implantation, each animal was fitted with a backpack carrying an external
transceiver coil and a battery-powered data acquisition system, and the dogs were allowed to freely
walk down a hallway. EMG recorded from each animal had very low noise and, in conjunction with
recorded video, clearly indicated swing/stance phases of gait.
This study demonstrates this design approach can be used to amplify and transmit muscle signals.
We are continuing development to reduce the size of implant to fit in a small, hermetically sealed
ceramic enclosure. This approach has the potential to substantially improve the control of prosthetic
limbs by providing simultaneous, multiple degree of freedom control, especially if used with
advanced prosthetic arms/hands, targeted muscle reinnervation patients, and recently developed
pattern recognition algorithms.
This work is supported by a National Institutes of Health SBIR Fast-Track grant (U44 NS067784).
The effect of subthalamic nucleus inhibition in the survival of
dopaminergic grafts in a rat model of Parkinson’s disease: An in vivo
platform for testing Smart Energy-Autonomous Micronode devices in
models of movement disorders
*Furlanetti LL1, Cordeiro KK1,3, Cordeiro JG1,3, Garcia J2, Winkler C2, Döbrössy M1, Nikkhah G1
Laboratory of Molecular Neurosurgery, University Hospital Freiburg, Germany
Department of Neurology, University Hospital Freiburg, Germany
Department of Neurosurgery, Federal University of Paraná, Brazil
Corresponding author’s email: [email protected]
Keywords: Parkinson’s disease, cell therapy, deep brain stimulation, subthalamic nucleus
Introduction: Parkinson´s disease (PD) is a neurodegenerative disease leading to progressive and disabling
deterioration of motor and cognitive skills. Current surgical treatment of inhibition of the subthalamic nucleus
(STN) by deep brain stimulation (DBS) can temporarily improve motor symptoms. Alternative regenerative
approaches are under research to restitute dopaminergic neurotransmission and offer a more extensive and
long lasting repair. However, sparse data are available concerning the combination of cell therapy and
neurotechnical modulation.
The aim of our project is to establish DBS in a rodent model of PD, and test whether the high frequency
stimulation within the basal ganglia can act synergistically with dopaminergic grafts in reversing functional
deficits. The rationale for the study comes from i.) pre-clinical and clinical experience showing that
dopaminergic grafts can influence function in PD or in models of PD; ii.) clinical experience showing DBS to
provide effective, long-lasting symptomatic relief to PD patients from diverse movement impairments; and
iii.) a pilot study showing the lesioning of the sub-thalamic nucleus to improve dopaminergic graft survival in
the rat model of PD. The poster presents our results from the pilot work, and introduces the follow-up
experimental design.
Method: Thirty-nine adult rats were rendered parkinsonian by unilateral injection of 6-OHDA neurotoxin into
the right medial forebrain bundle (MFB). The MFB lesioned animals were assigned into two groups, and one
group received an additional unilateral lesion of the STN by stereotactical injection of quinolinic acid in to the
STN. All animals were given primary striatal grafts of rat E14 ventral mesencephalic (VM) tissue into the
ipsilateral striatum. Drug-induced rotation, the Cylinder and the Stepping tests were performed to evaluate the
effect of each intervention. Post-mortem immunohistochemistry was carried out to assess the lesions and the
dopaminergic grafts.
Results: Functional recovery was observed in the behavioural tests in both groups after transplantation.
Survival of transplanted VM cells was observed in the striatum of both groups, however the histological
evaluation clearly revealed a positive effect of the STN suppression on the graft development.
Conclusion: Functional recovery and graft survival were observed after striatal transplantation, and improved
by means of STN inhibition. These findings suggest that cell therapy could be combined with other techniques
of STN suppression such as DBS. We are currently evaluating whether dopaminergic grafts and DBS can act
synergistically in the experimental model by assessing read-outs as graft survival and host re-innervation,
behavioural recovery (in freely moving animals during stimulation), as well as on neuronal spike activity in
striatum and motor cortex. (Figure 1) The in vivo platform of movement disorder will provide a model for
testing existing and novel Smart Energy-Autonomous Micronode (SEAM) devices as they emerge from the
Brain Links-Brain Tools consortium.
Fig. 1: (a) Stereotactic electrode placement in the right subthalamic nucleus (STN); (b) Local field potential
recording of the striatum and motor cortex; (c) Continuous high frequency stimulation of the STN in a grafted
freely moving animal model of Parkinson’s Disease.
Simulated prosthetic vision: object recognition and localization
Grégoire Denis, Marc J.-M. Macé*, Christophe Jouffrais.
IRIT, UMR5505, Université de Toulouse & CNRS - Toulouse
* CA: [email protected]
Blindness, Visual Prosthesis, Simulation, Object recognition, Object localization.
Currently, there is no efficient treatment against eye diseases such as retinitis pigmentosa (RP) or age
related macular degeneration (ARMD). They affect millions of people worldwide and may result in
blindness after a few years of evolution [1]. Within the last decades, a large collection of assistive
devices have been designed to assist people with visual impairment and enhance their autonomy.
However, assistive devices didn’t prove to be very useful in complex tasks such as navigation or
object recognition and grasping. In the meantime some research groups aim to restore vision via
neural interfaces, and have designed varied neuroprosthesis that elicit perceptible spots of light
(phosphenes). Today, the most advanced prostheses are implanted in the retina, and are based on two
different strategies. In the first one, an image is captured from a micro camera mounted on a pair of
goggles, then transmitted to a video processor and finally converted into electric pulses by an
electrode array implant. In the second one, an array of light sensors implanted in the retina directly
converts the incoming light into electrical signals [2]. Both devices have been designed to restore an
image of the whole field of view through a point-to-point display of phosphenes (sometimes called
“scoreboard” approach). The first clinical trials with arrays of 60 electrodes are encouraging:
implanted blind subjects are able to perceive simple visual stimuli in a highly contrasted and
controlled environment. However, more complex visuo-motor tasks, such as grasping an object
among other objects, are still very limited with these neuroprosthesis.
Material and methods
To overcome the limits of current visual prosthesis, we developed an alternative approach based on
object recognition and localization. Our hypothesis is that we may restore visual scene perception
and complex spatial behavior on the basis of a sparse representation of the surrounding space (a few
representative phosphenes only are highlighted). In this study we tested this hypothesis with a simple
object localization and grasping task. We developed a simulator of prosthetic vision (SPV) following
the recommendations published by Chen et al [3]. The SPV was made of a Head Mounted Display
with a stereo camera that captured the scene in real-time. We used a bio-inspired algorithm [4] to
perform real-time recognition and localization of target objects in the image. In this experiment,
twelve subjects were seated on a chair in front of a table. In each trial, the subjects were asked to
localize and grasp an object which was placed among nine other objects randomly spread over the
table. Four conditions were systematically assessed: (SC1) scoreboard approach with a simulated
6*10 electrode array, (SC2) scoreboard approach with a simulated 15*18 electrode array, (SC3)
scoreboard approach with a simulated 32*38 electrode array, and (LOC) localization approach with a
simulated 6*10 electrode array. Each subject performed 24 trials per condition. Performance was
measured in terms of time (time to grasp an object) and accuracy (% of correctly grasped objects). In
the SC conditions, the camera image was resized to fit with the number of electrodes available in
each condition (6*10, 15*18 & 32*38) and reduced to 8 grey levels. The luminance of each pixel in
this low resolution image was then used to display a round shape phosphene with a Gaussian profile.
Concerning the LOC condition, the aspect of the phosphene was the same as in the SC condition, but
we only displayed the nearest phosphene related to the localization of the target object in the image.
The average time to grasp the target object (correct trials only) was 22.8s (SE=4.8s) in SC1, 26.1s
(SE=3.6s) in SC2, 22.4s (SE=3.3s) in SC3, and 17.4s (SE=3.0s) in LOC. There were no significant
differences between any of these conditions (Friedman ANOVA, χ²=7.3, df=3, p=0.06). The
accuracy in SC1 was slightly above chance level (15.6%, SE=3.9% - Chance level was 10% correct),
and better for SC2 (37.5%, SE=9.6%). The accuracy was very good and comparable for the two
remaining conditions (SC3: 71.9%, SE=10.8%; LOC: 80.9%, SE=9.4%). The effect of the condition
on accuracy was significant (χ²=32.2, df=3, p < 0.001). Pairwise comparisons indicated that the
difference between SC1-SC3 and SC1-LOC was highly significant (p < 0.001). Another significant
difference was revealed between SC2 and LOC (p < 0.01) with a better accuracy in the localization
We showed that performance in object localization and grasping is very good in the LOC condition,
even with 60 electrodes only. Indeed it is similar to the performance reached with the classical
“scoreboard approach” but with 1216 electrodes (32*38). Currently implants with 1000+ electrodes
have been tested, but they didn’t prove to be more efficient. In fact, apparent resolution with 1500
electrodes seems comparable to apparent resolution with 6*10 electrodes, probably because of crosstalks between electrodes. Altogether, these results suggest that the effectiveness of the actual and
near future small electrode arrays may be enhanced if a contextual object localization algorithm was
included in the visual neuroprosthesis. In addition to object localization and grasping, the interleaved
artificial vision may subserve other perceptual [5] and visuo-motor tasks (pointing, reaching,
heading, etc.).
WHO, “Visual Impairment and blindness Fact Sheet N° 282,” World Health Organization,
E. Margalit, M. Maia, J. D. Weiland, R. J. Greenberg, G. Y. Fujii, G. Torres, D. V.
Piyathaisere, T. M. O’Hearn, W. Liu, G. Lazzi, G. Dagnelie, D. A. Scribner, E. De Juan, and
M. S. Humayun, “Retinal Prosthesis for the Blind.,” Survey of Ophthalmology, vol. 47, no. 4,
pp. 335–356, 2002.
S. C. Chen, G. J. Suaning, J. W. Morley, and N. H. Lovell, “Simulating prosthetic vision: I.
Visual models of phosphenes.,” Vision Research, vol. 49, no. 12, pp. 1493–1506, Jun. 2009.
F. Dramas, S. Thorpe, and C. Jouffrais, “Artificial Vision For The Blind: A Bio-Inspired
Algorithm For Objects And Obstacles Detection,” International Journal of Image and
Graphics, vol. 10, no. 4, pp. 531–544, 2010.
R. Parlouar, F. Dramas, M. J.-M. Macé, and C. Jouffrais, “Assistive device for the blind based
on object recognition: an Application to Identify Currency Bills,” in SIGACCESS conference
on Computers and accessibility, 2009, pp. 227–228.
Spike detection during electrical stimulation in a population of retinal
ganglion cells
F.Helmhold*, M.Eickenscheidt, G.Zeck
Neurochip Research Group,
Natural and Medical Sciences Institute at the University of Tuebingen, Tuebingen, Germany
* [email protected]
Neurochip, Retina, CMOS, Electrical Stimulation
The performance of electrical neural stimulation is often assessed using the recording of a single cell
only. However, each stimulus may excite more than one cell in the vicinity of the stimulation
electrode but also at remote distances due to axonal stimulation [1].
The mammalian retina with its layered structure is a good model tissue to study the electrical
activation of inter-neurons and projection neurons [2, 3]. So far it has been proven difficult to
estimate electrical activation of multiple cells during stimulus presentation due to the so-called
stimulus artifact. Here we present experiments which avoid this limitation and allow for performance
testing of various neural stimulators.
A high-density CMOS-based microelectrode array comprising 16384 recording sites (pitch of 7.4
µm) [4] is used to record current profiles at a frame rate of 12 kHz. Whole-mount retinas were
interfaced to the array as previously described [5]. Here electrical stimulation is performed using a
metal electrode by applying charge balanced biphasic constant current pulses. The stimulation
electrode is placed in a subretinal configuration (Fig. 1 A). Repetitive stimulations allow a spatial
stimulus profile characterization and artifact subtraction. Extracellular waveforms are assigned to
corresponding retinal ganglion cells following a recently published sorting algorithm, taking
advantage of the high density of recording sites [4].
Guinea pig and rat retina were stimulated using tungsten electrodes placed on the photoreceptor
layer. The amplitude of charge balanced pulses was varied between 1 and 25µA and single phase
durations between 100µs and 5ms (Fig. 1B i). The capacitive recording electrode (field effect
transistor) does not saturate at the applied stimulation currents thus allowing for the detection of
spikes during the stimulation pulse (Fig. 1B ii).
Stimulation at a given amplitude and position (Fig. 1B iii white circle) elicit reliable spikes in
ganglion cells (Fig. 1B ii) We detect evoked activity in several ganglion cells simultaneously, which
are stimulated by the same pulse at a distance up to 400µm from the stimulation center. Cells exhibit
thresholds dependent on the distance from the stimulation center. The thresholds increase from 9µA
at d=60µm up to 22µA at d=400µm. By changing the position of the stimulus the dependency
between threshold and distance can be studied in detail.
Fig. 1 Experimental procedure. (A) The high-density sensor array is capacitivly interfaced to neural tissue (retina) through an inert
oxide. Retinal Ganglion Cells (RGC) are in close contact to the sensing transistors (sketch). A neuronal simulator is placed in
subretinal configuration and biphasic current pulses are applied (B, i) A single transistor of the array detects a large stimulus artifact
overlain by action potentials of one RGC (B, ii) The cell activity is also seen by neighboring transistors which can be used to identify
the cell response (B, iii). The sensor trace shown in (B, ii) is marked with a white rectangle on dark background. The stimulation
location is marked with a white circle.
Our system can be used to investigate the performance of various neuroprosthetic devices in terms of
signal distribution and tissue response.
This work was funded by a grant of the German Ministry for Education and Research (BMBF FKZ:
[1] G.Beit-Yaakov, D. Raz-Prag, Y. Hanein “Highly non-localized retinal activation verified with
MEA and Ca imaging techniques” Proc. of the 8th MEA Meeting 2012, Reutlingen, Germany
[2] Stett A, Barth W, Weiss S, Haemmerle H, and Zrenner E. (2000) Electrical multisite stimulation
of the isolated chicken retina. Vision Research 40: 1785-1795.
[3] Eickenscheidt M., Jenkner M., Thewes R., Fromherz P., Zeck G. (2012) „Electrical Stimulation
of Retinal Neurons in Epiretinal and Subretinal Configuration using a Multi-Capacitor-Array”
J.Neurophys. 107(10):2742-55.
[4] Lambacher A, Vitzthum V, Zeitler R, Eickenscheidt M, Eversmann B, Thewes R, and Fromherz
P. (2011) „Identifying firing mammalian neurons in networks with high-resolution multi-transistor
array (MTA).“ Applied Physics A 102: 1-11.
[5] Zeck G, Lambacher A, Fromherz P (2011) “Axonal Transmission in the Retina Introduces a
Small Dispersion of Relative Timing in the Ganglion Cell Population Response.” PLoS ONE 6(6):
Electrically Evoked Responses of Retinal Ganglion Cells in Wild-type
and Rd10 Mouse Retinas.
Archana Jalligampala 1,3,4, Daniel.L Rathbun 1,2,3*, Eberhart Zrenner 1,2,3
Institute for Ophthalmic Research, University of Tübingen, Germany.
Bernstein Center for Computational Neuroscience Tübingen, Germany.
Centre for Integrative Neuroscience, Tübingen, Germany.
Max Planck Graduate School of Neural and Behavioral Sciences, Tübingen, Germany.
*Corresponding author: [email protected]
Epiretinal stimulation, rd10, RGC, flex-MEA
Objectives: Over the years, retinal implants have been developed to restore limited functional vision
in patients blinded by outer retinal diseases like retinitis pigmentosa (RP) and age–related macular
degeneration through electrical stimulation of the surviving neurons. The most extensively
characterized animal model for human RP is the rd1 mouse. However, the recently identified rd10
mouse, which has a relatively delayed onset and slower progression of degeneration may be a more
appropriate model. To support ongoing efforts to optimize prosthetic retinal stimulation, stimulation
paradigms need to be established for this new mouse line. Here we investigate retinal ganglion cell
(RGC) responses to different stimulation paradigms in adult wild-type (wt) and rd10 mice.
Methods: RGC spiking responses were recorded in vitro from patches of wt and rd10 retina
epiretinally, using a planar multi-electrode array (MEA, 60 electrodes, 200µm interelectrode
distance, 30µm diameter, 8X8 layout; MultiChannel Systems, Reutlingen, Germany). Prior to
epiretinal electrical stimulation, spontaneous activity was recorded. The stimuli were applied to the
retina via one of the 60 electrodes on the MEA while the other electrodes recorded electricallyevoked responses (MC-Stim, and MC-Rack MultiChannel Systems). Stimuli consisted of squarewave, monophasic (both cathodic and anodic) voltage pulses with randomized pulse durations.
Duration randomization and incremental voltage blocks were employed to compensate for recording
instability-induced biases that we observed in preliminary experiments. The stored data were
processed & analyzed offline using spike sorting software (Offline Sorter & NeuroExplorer, Plexon
Inc, TX) and custom Matlab scripts (The Mathworks, Natick, MA) to generate rastergrams, peristimulus time histograms and response curves. The same experimental design and methods were
applied during subretinal stimulation in which a flex-MEA (36 electrodes, 300µm interelectrode
distance, 30µm diameter,6X6 layout; MultiChannel Systems, Reutlingen, Germany) was placed on
the subretinal side to deliver electrical stimulation while the planar MEA recorded RGC responses.
Results: In agreement with recent reports, spontaneous activity was higher and more oscillatory in
rd10 retina than in the wild-type (Goo 2011). Under epiretinal stimulation, the wt retina
demonstrated voltage and duration requirements that are in agreement with previously published
reports. In both wt and rd10 cells, a dependence of the response on duration was typically seen only
for a few transitional voltages. Both above and below these voltages the influence of duration was
minimal. We find no significant differences in stimulus threshold between wt and rd10 retina (Fig 1).
Previous studies remain unreconciled on whether thresholds are higher or lower in degenerated
retina, likely reflecting influences of mutant strain, species, and age (Chan 2011; Jensen 2008). A
subset of cells demonstrated an asymmetric preference for either negative or positive voltage pulses.
Additionally, ganglion cell responsiveness decreased with increased interelectrode distance to around
800µm from the site of stimulation. Furthermore, some cells in both wt and rd10 retina showed an
initial suppression of ongoing activity upon epiretinal stimulation. Finally, we have developed a
method for subretinal stimulation in which a flex-MEA is pressed onto the photoreceptor side of a
retina already mounted ganglion cell side down onto a standard MEA chip. Using this ‘sandwich’
approach, we have made initial epiretinal recordings with simultaneous subretinal stimulation in wt
Conclusions: Our preliminary findings present one of the first examinations of electrical stimulation
in rd10 retina. Based on these findings, we propose tentative stimulation parameters appropriate for
activation of rd10 retina in our continued development of more efficient stimulation protocols for the
Tübingen retinal prosthesis.
Earlier versions of this work has previously been presented at the MEA Meeting 2012, Reutlingen,
Germany; Eye and the Chip 2012,Detroit,MI ,USA and Bernstein Conference 2012,Munich,
Fig 1: Stimulus strength/duration response curves. (A & B) Threshold voltages for constant-duration response curves
(C & D) Threshold durations for constant-voltage response curves. Threshold was defined as the average baseline firing
rate + 2*variance.
Modification of Flexible Probe by Electrochemical Etching to Reduce
Paula Klimach1, Yijing Xie2, Ulrich Hofmann3
Institute for Signal Processing, University of Lübeck, 23538 Lübeck, Germany
Graduate School for Computing in Medicine and Life Sciences, University of Lübeck, 23538 Lübeck, Germany
Neuroelectronic Systems, Department of Neurosurgery, University Medical Center Freiburg, 79108 Freiburg, Germany
Keywords (up to 5)
Flexible electrodes, flexible probes, surface modification,, electrochemical etching
ABSTRACT (up to 1000 words, no longer than 2 pages)
Flexible probes are thought to be the state of the art method for chronic intracortical neural recording
and stimulation. As microelectrode sites on such a probe become smaller their impedance increases.
Increased impedance reduces the quality of the signal recorded and thus produces a conflict between
the desire for a high quality signal and smaller and more selective sites. It is therefore desirable to
reduce the probe impedance in order to increase signal quality while maintaining the size of the site.
Electrochemical etching was the method chosen to modify the surface of recording sites on a flexible
probe because it is a simple and cost effective method of surface modification.
The aim of the experiment described here was to determine the parameters necessary to etch the
microelectrodes on a flexible probe and achieve the largest reduction in the impedance. The study
focused on the amount of current applied on the system and the amount of time required. The
investigated probe was a polymide based flexible probe with 16 gold contacts. Of the 16
microelectrodes 4 were stimulating sites and 12 recording sites. The stimulation sites have a
lithographical produced surface area of 3.6x10-5cm2 and the recording sites of 4x10-6 cm2.
Impedance measurements were taken for the contact sites in a 0.9% NaCl electrolyte using an
LCR800 impedance meter. The setup consisted of a three electrode cell with an Ag/AgCl reference
electrode, platinum wire counter electrode, and the flexible probe working electrode. The current was
applied by the Keithley 6221 precision current source. At each experimental step a current was
applied to the system for 10 seconds after which the impedance was measured. There were nine
steps, therefore a cumulative application of current for a period of 90 seconds. Measured
temperatures of the saline were between 24oC to 28oC.
The setup above was used with different current values. The result of the experiments was to
determine the greatest difference between the original impedance and the decreased impedance. The
current that lowered the impedance the most was 0.4 μA. Initial impedance values of about 3490kΩ
at 1kHz were reduced to an average impedance of 1261kΩ. The set current ranges from 0.28 μA up
to 100 μA. Current values lower than 0.28 μA do not create a difference in impedance measurement;
currents larger than 0.4 μA create a difference but the decreased impedance value is not as large as at
0.4 μA..
We conclude that a simple galvanic etching process can improve the impedance of lithographically
deposited gold microelectrodes by 21%.
What Blind Retinitis Pigmentosa Patients Can See when Using the New Subretinal Wireless Implant
E. Zrenner , K.U. Bartz-Schmidt , A. Braun , A. Bruckmann , F. Gekeler , U. Greppmaier , S. Hipp , G.
Hoertdoerfer , C. Kernstock , A. Kusnyerik , H. Sachs , K. Stingl .
Institute for Ophthalmic Research, Centre for Ophthalmology, Tuebingen, Germany;
Retina Implant AG, Reutlingen, Germany;
Mobility Training, Tuebingen, Germany;
Semmelweis University, Budapest, Hungary;
Städtisches Klinikum Dresden-Friedrichstadt, Dresden, Germany.
Purpose: Subretinal microphotodiode arrays with 1.500 pixels are able to restore vision up to reading
capability (Zrenner et al . Proc. R. Soc. B 2011, 278: 1489ff). Here we report about the first phase of a
multicenter trial with the wireless implant Alpha-IMS (Retina Implant AG, Tübingen, Germany).
Methods: Each of the 1500 subfoveal photodiodes within a 11 by 11 deg field controls an amplifier that,
depending on the strength of the light ejects currents onto bipolar cells via an electrode (Stett et al. Vision
Research 2000). Power and control signals are supplied by inductivity via a subdermal retroauricular coil from
which a subdermal cable leads to the eye ball. The new device was implanted in ten patients since 2010
(average age 45.95±7.9; 5 males, 4 females). Function was tested by four procedures: 1. Monitor-based
standardized tests with controlled conditions for testing light perception threshold, light localization and
movement (Wilke et al. 2007), as well as grating acuity and Landolt C-rings (2 or 4 AFC); 2. Recognition tasks
at a table setting with table ware and geometric objects; 3. Reading of letters; 4. Outdoor activity reported by
the patients; all test results were controlled with power switched off.
Results: In all ten patients the chip was at the desired subfoveal position except in two patients where it was
slightly parafoveal. Proper chip function was proven by measuring chip output via electroretinography. All
patients were able to perform the function tests except one where a loss of inner retina function was observed
after surgery. Results in all other patients were: light perception: 9/9, light localization: 8/9; motion recognition
5/9; grating resolution 8/9 (up to 3,3 cycle/degree); Landolt C rings 3/9 (up to 0,036); recognition of geometric
objects 8/9; recognition of objects in table setup 8/9; Letter reading 4/9; clock hands reading 3/9; grey scale
differentiation 6/9; improved outdoor mobility and activity 5/9.
Patients reported numerous beneficial visual experiences in daily life with regained recognition of unknown
objects, recognition of facial or clothes’ characteristics, vapor of jet airplane on the sky, moving objects in
nature and traffic, improved self-sustaining actions (recognition of doors, door handles), recognition of small
objects (glasses, telephone, stapler, washing basin, even dices and numbers of dots on dices), improved
mobility, walking along street lanterns during night, recognizing car head and rear lights.
Conclusions: Careful psychophysical testing and patient daily-life reports show that the wireless Alpha-IMS
implant restores useful visual abilities in blind RP-patients. Subretinal surgery for positioning chips subfoveally
is safe and the multicenter part of the study has been started in 3 European countries.
Fly UP