Chapter 3 Flexible Brain Surface Grid Electrode Array
3.4 In-vivo Recording of Auditory Response in Rat
Fig. 3.2 Fabricated grid electrode array Fig. 3.3 Impedance characterization of single electrode
3.4 In-vivo Recording of Auditory Response in Rat
To demonstrate the functionality of fabricated gird electrode array, in-vivo experiment on auditory stimulation response recording is induced. When an authorized rat receives specialized frequency/magnitude sound stimulation, electrical response can be observed on a localized auditory area of cortex. It can be seen as neural activity happens due to the ears receive the sound and transfer into neural signals into brain and appear in auditory sensory brain area. By using the flexible grid electrode array, the interested area of brain is covered and recorded. Therefore, localized electrical response spot could be detected and characterized for functional mapping and event evaluation.
Sprague Dawley rats, aged 10–12 week old and weighting 280–350 g body weight, were used in the experiments. The animals were kept in a room under a 12:12-hr light-dark cycle with food and water provided ad libitum. All surgical and experimental procedures were reviewed and approved by the Animal Ethics Committee of the National Chiao-Tung University. The rats were anesthetized with urethane 2.0 g/kg body-weight (b.w.) intraperitoneal (i.p.). Subsequently, it was placed in a standard stereotaxic apparatus.
Proposed grid electrode array are implanted over the area of auditory cortex. Fig. 3.4 (A) shows the implanted location on the rat brain [57], and the practical implantation optical photograph is presented in Fig. 3.4 (B), which shows that the presented grid electrode displayed a flexible property therefore greatly fit to the exposed brain cortex. Fig. 3.4 (C) indicates the close view of the electrode array contact area. Note that only the electrode covered area is exposed to dura layer, the rest area is skull. Fig. 3.4 (D) illustrates the channel number with respect to the location in electrode array, where ch-14, ch-15, ch-16 acts as ground and ch-1 is used as the reference channel.
The audio stimulator use signal generator, amplifier, audio amplifier and programmable attenuator (TDT-RZ5, TDT-PA5, TDT-RP2.1, TDT-SA1). The stimulate signals are 1kHz, 2kHz, 4kHz and 9kHz in frequency (in the auditory range of rat, 250-60kHz) with 25ms
Fig. 3.4 Electrode implantation location, rat brain anatomy is adopted from reference [58]
duration, 2.5ms linear rise/fall time and 75dB sound pressure level (SPL). Every test sound is repeated for 100 times with 1 second period. Additionally, the speaker output is calibrated by a precise microphone (B&K 4149) before experiment.
Fig. 3.5 shows an example of recorded 16 channel signal in 1 second time frame. The upper green line in the figure illustrates the stimulate signal. The auditory response is then recorded and averaged to explicate the even-related evidence by software.
Fig. 3.5 16-channel recoding in 1 second time frame
Fig. 3.6 shows the time-magnitude plot of the averaged result of the 16 channel auditory ECoG response: (A) 1 kHz 75 dB SPL (B) 2 kHz 75 dB SPL (C) 4 kHz 75 dB SPL (D) 9 kHz 75 dB SPL. The 4 by 4 matrix boxes indicates the related electrode location on the brain. Vertical line in each box denotes the stimulation trigger marker. Four electrodes maker G (ground) and R (reference) are used as ground electrodes and reference electrodes.
Fig. 3.7 presents the time-Frequency plot: (A) 1 kHz 75 dB SPL (B) 2 kHz 75 dB SPL (C) 4 kHz 75 dB SPL (D) 9 kHz 75 dB SPL. Related frequency distribution versus different electrode location can be observed in the results.
Fig. 3.6 Time-magnitude plot of the averaged 16 channel auditory ECoG response (A) 1 kHz 75 dB SPL (B) 2 kHz 75 dB SPL (C) 4 kHz 75 dB SPL (D) 9 kHz 75 dB SPL.
Fig. 3.7 Time-frequency plot of the averaged 16 channel auditory ECoG response (A) 1 kHz 75 dB SPL (B) 2 kHz 75 dB SPL (C) 4 kHz 75 dB SPL (D) 9 kHz 75 dB SPL.
By comparing the recorded ECoG intensity, 4 kHz stimulation shows strongest response, while 2 kHz shows less response and 1 kHz, 9 kHz have weakest response signal. Measured results shows that localized neurons discharge phenomenon (gamma activities, 30 – 200Hz) are recorded by the presented grid electrode array. According to Fig. 3.6 and Fig. 3.7, three different auditory areas are covered by the grid electrode due to the different pattern observation. Group 1 includes 13, 11, 5, 3, Group 2 includes 8, 6, 10, 12, 2, 4, and 9, 7
belongs to the other group. Experiment results shows that recorded evoked potential on auditory cortex is about ±70uV, and the minimal effective sound stimulation magnitude is 20 dB SPL.
3.5 Summary
In this chapter, the implementation and characterization of a MEMS-based flexible grid electrode array utilizing parylene-C as substrate was presented for ECoG measurement applications. An un-symmetric sandwich-type structure consists of two Parylene layer (10um substrate and 1um isolation layer) and one Platinum layer was successfully fabricated. Comparing to previous works that using symmetric sandwich-type with thick isolation structure (polyimide, >20um), proposed method enhanced the adhesion property on brain cortex surface with great flexibility. Additionally, previous works suffer from the level difference between sensing electrode and passivation surface may need another electroplating post process high curing temperature (200-400ºC). In this work, simplified process under room temperature and superior properties were achieved without complex and high-temperature processes.
In-vivo experiments demonstrate the recording capability of the proposed grid electrode
array. Recorded auditory evoked potentials (AEPs) is ±70μV with 20dB SPL minimum sound level under general anesthesia, as well as the localized neurons discharge phenomenon (GAMA Activity, 30Hz~500Hz). The measurement result shows that the presented electrode array can cover the most of brain auditory cortex surface area, and distinguish specific signal characteristics between different electrode sites. Proposed grid array is looking forward to be used in awake animals in future works, which helps the studies on neurons degeneration, the mechanism of brain disease, and the development of the ECoG controlled BCI interface as well.Chapter 4
Three Dimensional Neural Probe Array
4.1 3-Dimensional Electrophysiology Recording of Biological Cells
In recent years, advance micromachined/assembled micro probe arrays with electrical recording ability have come to play an essential role in exploring central neural systems.
Simultaneous observation of a larger number of cell activities has become the general requirement to understand the nervous system [58]. Advances in neuroscience and neuroprosthetics now require microelectrode arrays that are able to access numerous neurons simultaneously with high spatial resolution [59]. Recording of the extracellular action potentials has been accomplished by surgically implanting neural probes into the target neurons of interest, which resulted from neural activities. Probes that could insert a large number of recording sites into neural tissues with minimal tissue damage are therefore needed. Also, the design of the probe arrays should be optimized for an experimental purpose that an electrode diameter of a few micrometers could support single-unit recording [60].
The traditional micro probes, which are made from insulated metal wires and glass micropipettes, cannot provide simultaneously multi-channel recording. The main reason is that the traditional devices function as only a single site on a single probe shaft. Some previous studies have improved the problem by thin-film lithography-based micromachining techniques since 1960s.
High-density probe arrays yielded insights into the organization and function of the neural system [61]. Silicon [62], glass [63], polymer [64] and sapphire [65] substrates have been employed as thin-film electrode probe planks. The thin-film silicon micro probe was developed many years ago for neuroscience and neural prostheses [66]. It has also been widely characterized electrically [67] and mechanically [68] for probe scaling [69], insertion force [70], tissue strain [71] and chronic brain responses [72]. The studies mentioned above provide detailed multi-channel recordings along a single plane, but lacked of full cell activity information in 3-D space [56].
To access the full cell activity that originates in the target tissue, three dimensional microprobe arrays are strongly required with precisely controlled dimensions and front-end circuitry compatibility. In other words, to achieve detailed studies of neural networks and
implementation of neural prostheses, we need to access three-dimensional volumes of tissue with three dimensional distributed recording sites. In modern neural system researches, 3D microprobe array allows the recording and mapping of the neural signal network and interconnections among the 3D brain structure. The recording and mapping would be impossible to achieve by using 2-D planar arrays [56].
Creating 3-D arrays by the assembly of 2-D parts is now the most popular method to construct a 3-D structure [59] [73-77]. The 2-D parts usually include 2-D arrays, vertical spacers and supporting platform. The supporting platform acts as a substrate, and the vertical spacers are erected on the supporting platform by tethers, joints and snap fasteners.
The spacers fixed the 2-D arrays vertically on the supporting platform, and made the probe shafts pass through the holes of the supporting platform. The full 3-D structure is therefore like a PC motherboard. Additionally, active circuitry for signal processing can be designed and fabricated in the back-end of the 2-D arrays to achieve system integration. A unique handling method was developed in [76] for a dual-side, ultra-thin silicon substrate process to fabricate thin probe shafts without using doping etching stop technique. Moreover, stacked probes and PCBs by anisotropic conductive film create the connection for the dual-side wire routing and 3D structure. Therefore, each side of probe can be wired out separately. An alternative solution provided in [77] integrated the silicon probe with flexible ribbon cables by using thermosonic bonded gold bump. Also, a platform with bays and gold clips is designed to connect with probes, which results in an impressive 3D device. The comparison of three-dimensional microprobe arrays with some major design parameters is shown in Table 4.1. However, the studies mentioned above neglect the importance of smaller opening for surgery implantation. Smaller opening of skull can reduce the implantation damage to the subject, prevent the rise of brain pressure, and decrease the infection probability of the wound.
Although previous work creating 3-D arrays by assembly of 2-D arrays successfully achieves high electrode density by packaging active probes onto the supporting platform with some micromechanical packaging technique, some problems still exist. First, previous approaches that use 2-D silicon probes to form full 3-D arrays required complex schemes for assembling submillimeter parts [74]. The main problem of such techniques is that the parts (spacers and supporting platform) were all assembled in orthogonal planes. Thus, perpendicular connectors for interconnections between orthogonal planes were required for signal transmission. Ultrasonic bonding [74] and vertical snap fasteners [59] have been proposed for perpendicular transfer pads, but they suffered from complex assembly steps and precise alignment equipment for 3-D assembly. For example, precise alignment was required to make probe shafts pass through holes of the supporting platform and steady the probe onto the vertical spacers without damage during the assembly process. Second, the probe arrays were fixed only by the perpendicular bonding pads and the tenons. Low structure strength can cause stability problem in implantation. Third, the rooms between the
spacers and the 2-D probes were wasted. The volume of a 3-D structure increases rapidly when increasing the number of 2-D probes. Notably, the supporting platform also defines the minimal surgery opening requirement for implantation. Signal processing chips can either be made in the backend of the 2D arrays or mounted onto the supporting platform to achieve system integration. The former currently provides only low-end circuit processes, and the latter results in tremendously increased volume size and surgical opening for implantation. In fact, fabrication of probe array structures with advanced CMOS process in the same wafer can be a large cost because the probe area may usually much larger then the area of the backend.
Table 4.1 Comparison of 3D microprobe arrays with some major design parameters
Reference [9-10] [77-80] [81] [59, 73-75] [77] [76]
Substrate Si Epoxy,
Polyimide Polyimide/Nickel Si Si Si
Dimension 3D 3D 3D 3D 3D 3D
To improve the problems described above, this work reports a new stacking method for fabricating 3-D neural probe arrays. In this study, the 3-D orthogonal interconnection was replaced with 2-D wire bonding by the present stacking method, and the perpendicular bonding and snap fasteners which were used in previous work were no longer needed.
Compared to previous work, this new stacking method can also provide reliable structure strength. ASIC chips can be substituted for spacers to increase the system integration and volume usage efficiency as well. Additionally, an anti-overflow design based on the capillary principle was exploited to avoid gel overflow onto proximate bonding pad during 3-D array assembly.
4.2 Stacked Multichip Structure Design
A new stacking method to produce three-dimensional neural probe arrays is presented in this work. This method creates 3-D probe arrays by assembling 2-D arrays and spacers layer by layer, as shown in Fig. 4.1. For a 4 × 4 3-D array, four 2-D arrays (gray color) with four probes in each array and three spacers (yellow color) were required. Compared to exising three-dimensional neural probe designs, the present stacking method improved the inconvenient assembly steps which include orthogonal assembly and perpendicular connection techniques. In the stacking method, the shapes of each 2-D arrays were carefully designed so they can be wire-bonded individually with different height levels. Spacers with an anti-overflow mechanism were also proposed in this paper. The present anti-overflow mechanism can also be realized on 2-D arrays if active circuit chips are used as spacers.
Also, the thickness of the spacer determined the spacing between two 2-D arrays. Each planar 2-D array, electrode sites, inter-connection routing and bonding pads were located in the same plane. The bonding pads were arranged on the different sides of four 2-D probe arrays for wire bonding. Therefore, each 2-D array can be wire-bonded individually and the 3-D perpendicular bonding pads used in previous work are no longer needed.
Fig. 4.1 The schematic of stacking a 4 × 4 3-D microprobe array
As shown in Fig. 4.2 (A), presented design needs smaller skull opening for insertion comparing with previous works. In previous works, the area of supporting platform defines the minimal surgery opening on the skull for implantation. In this design, the area needed
for insertion is almost the same as the recording target area. Presented method also has advantages of easy assembly process and high structure strength because the vertical connecters, tenons and joints that used in previous works are no longer needed. Moreover, the spacers give the possibility for multichip circuit integration by replacing them by circuitry chips.
By replacing spacers with active signal processing circuitry chips, the function of the 3-D neural probe array can be enhanced. For example, a conceptual representation of the stacked-multichip neural sensing interface is shown in Fig. 4.2 (B). For 64 recording sites sensing purpose, 4 planar arrays (ARRAYn) which consist of 4 probe shafts with 4 recording sites on each probe shaft are stacked layer by layer. For each planar array ARRAYn, there is a 16-channel neural amplifier (AFEn) mounted onto itself by flip-chip bonding for amplification. Therefore, signal conditioning provided by the AFEn can achieve reasonably high signal-to-noise ration before the neural signal being transferred into ADCs via the bonding wire. Also, the 16-channel neural amplifier has a time-division series data output to reduce the number of I/O, that is, the amount of wire-bonding.
Fig. 4.2 (A) Comparison between presented stacking method and previous works using planar arrays vertically connected onto the supporting platform for constructing a 3D neural
recording. (B) Conceptual representation of the multichip neural interface
The lengths of the backend of each planar array are different therefore allow each layer of planar array can be wire-bonded individually with different level of height. Since that the size of each AFEn is always the same, there are three spacers therefore needed when stacking 4 planar arrays and 4 neural amplifier chips into a 3D neural recording device as shown in Fig. 4.2 (B). The ADCs is mounted on the supporting platform for signal digitalization and controlled by the micro-control unit (MCU). The MCU also controls the synchronization between AFEs and ADCs, and the data transmission via wireless RF modules. Related wireless data/power transmission modules and passive devices can also be mounted onto the supporting platform. Finally, biocompatible material is employed to encapsulate the implantable device for biocompatibility and electrical isolation.
Briefly speaking, in IC manufacture, larger area means higher cost. When the active
circuitry is fabricated with probe shafts in the back-end of the 2-D array, larger wafer areas are required. In the stacking design, spacers with circuitry and 2-D arrays were fabricated individually. Therefore, the stacking method can reduce the cost for circuitry integration and increase the design flexibility when modification of probes/circuitries is required for different applications. Besides, comparing with previous work, the volume usage efficiency was increased because there were no waste rooms between arrays and spacers. In short, the advantages of using active circuit chips as spacers include reducing the cost of circuitry integration, increasing the flexibility of the design and increasing the volume usage efficiency.
4.3 Fabrication, Assembly and Characterization
The key component in present stacked 3D structure is the planar neural probe arrays, which were made with multiple bio-sensing sites for neural recording [82]. The fabrication steps of the planar array are displayed in Fig. 4.3 (A) and briefly described as follows: 250nm-thick nitride deposition on 200μm-thick silicon wafer for electrical isolation. Then, 300nm/30nm-thick Pt/Ti layer was deposited and patterned by lift-off for wire interconnects on the probe shaft, following with 500nm-thick nitride deposition for encapsulation.
Electrode sites and wire-bonding pads were defined by RIE. Then, 2μm-thick SiO2 was deposited by PECVD to protect the probe structure. Finally, shape of probe array was defined and released by DRIE. Fig. 4.3 (B)-(C) display the fabrication results of the planar arrays.
Another similar fabrication process which using polyimide (PI) and Cr/Au as isolation and conduction layer is also developed. The thickness of PI and Cr/Au are 3um and 1um, following with 3um-thick, electroformed Au as electrode site and bonding pad material. The Au layer is somewhat over-electroformed to ensure that the electrode was in contact with the neural tissue while implantation. Also, the final shape of the planar array was defined and released by DRIE.
When the stacking method is used to construct the 3-D neural probe arrays, the overflow adhesion gel or glue between the stacking layers may cover the proximate bonding pads and make them ineffective. Using less gel may reduce the overflow problem, but reduce the adherent strength. To solve the overflow problem of the gel, an anti-flow mechanism design was applied in the stacking method. The anti-overflow mechanism was accomplished by creating a through-silicon-via around the edges of the spacers. It uses capillary action force to prevent the gel from overflowing to the bonding pads. The mechanism functions in the following condition: when the stacking process starts, the combined parts compress the adhesion gel and force it to flow around. The flowing glue will fill the via by capillary action as it passes the via. Therefore, there is no redundant glue covering the proximate bonding pads.
Fig. 4.3 (A)Fabrication steps of the planar array (B) FabricatedProbe tip. The tapered tip angle is about 23° (C) Fabricated parts on a one cent coin (D) Tip and electrode The radius of the via was one of the major design parameter in preventing overflow. The formula is given by the well-known capillary action principle [83] with definition of the liquid-air surface tension, contact angle, density of the liquid, acceleration due to gravity, the height of the liquid column and the radius of the via. In this case, the maximal height of the liquid column is the thickness of the spacer (250 μm), and the liquid-air surface tension is 0.033 N/m [84], contact angle is 70° [85], density is 2,000 kg/m3 and gravity acceleration is 9.8 m/s2. The capillary action principle gives the radius of the via a theoretical result of 4,600 μm, which was even larger than the size of the spacer. In fact, there is some limitations should be put into consideration. For instance, the limited volume of glue, the viscosity of glue, the friction force between glue-substrate interface and the capillary force
Fig. 4.3 (A)Fabrication steps of the planar array (B) FabricatedProbe tip. The tapered tip angle is about 23° (C) Fabricated parts on a one cent coin (D) Tip and electrode The radius of the via was one of the major design parameter in preventing overflow. The formula is given by the well-known capillary action principle [83] with definition of the liquid-air surface tension, contact angle, density of the liquid, acceleration due to gravity, the height of the liquid column and the radius of the via. In this case, the maximal height of the liquid column is the thickness of the spacer (250 μm), and the liquid-air surface tension is 0.033 N/m [84], contact angle is 70° [85], density is 2,000 kg/m3 and gravity acceleration is 9.8 m/s2. The capillary action principle gives the radius of the via a theoretical result of 4,600 μm, which was even larger than the size of the spacer. In fact, there is some limitations should be put into consideration. For instance, the limited volume of glue, the viscosity of glue, the friction force between glue-substrate interface and the capillary force