Several groups worldwide are now working towards implantable wireless active microelectronic neural interfaces [10
], with recent progress reflected by papers published in this issue. The concept of an implantable wireless platform, without any percutaneous skin-puncturing elements, demands the implementation of a heterogeneous, active microelectronics platform. At a minimum, the implanted “microsystem” requires in-situ integration of ultra-low-power microelectronic ASICs with the cortical microelectrode neural probes, and must provide broadband telemetry and a means to deliver power wirelessly to the active implanted components. In our laboratory, a 16-channel version of a fully implantable microsystem, whose design and performance is summarized in the following sections, has been implemented and tested on the benchtop and used in initial animal tests. Our microsystem is a single-unit construct () where analog and digital chips are integrated on a flexible substrate together with a low threshold, infrared semiconductor diode laser to transmit the digitized neural signals through the skin. Power and clocking are delivered to the system via inductive coupling [13
], but the system can also be configured to be powered optically using a high efficiency photovoltaic energy converter [12
Fig. 1 Schematics of the ‘dual panel’ brain implantable microsystem featuring an active brain sensor (microelectrode array integrated to amplifier IC) in the cortical unit, and hybrid A/D, control, and RF-IR(Infrared) telemetry in the cranial (more ...)
The present use of a 16-electrode sensor as a development model of our system reduces its complexity, fragility and cost. This allows us to test telemetry and power transfer, package design, and surgical adaptability more thoroughly and quickly. However, all aspects of the system have been designed to be scalable in two senses: first, the amplifiers and ADC systems scale nearly linearly in power with increasing numbers of electrodes. We have already established that our inductive coupling loop is capable of supplying the additional power and the preamplifiers (45 microwatt per electrode) are sufficiently low power that their thermal load on the cortex will still be acceptable. The infrared telemetry has so far demonstrated capacity for 32 channels in vivo and is scalable to 100 channels. The digital processing is adaptable in its current form to any number of channels as long as the clock rate, limited by the telemetry channel, is sufficient to carry the data.
The second sense of scalability refers to the ability to place two or more separate cortical sensors in the cortex so that as more channels are available, they can be used to acquire data from different brain areas simultaneously. (One example is to implant one array in the motor cortex and a second in the parietal cortex). This is possible because we separate the signal processing, power, and telemetry subsystems from the electrodes and their analog amplification and multiplexing. Channel addressing and ADC conversion have been designed to merge two sets of data into a single output stream at no increase in power, alteration in chip design, and no change in the number of wires to each array.
The rationale for our choice of the particular spatial distribution of the passive (cortical) neural sensor, the active signal acquisition electronics, the digital control components, and the telemetry and power elements, is based on convergence of considerations of microelectronic device engineering, thermal loading, neurophysiology, anatomy, and surgical constraints. At one extreme of such possible layouts is a construct with all the microelectronic components physically integrated directly onto the multielectrode neural sensor. Such a monolithic “brain button” would be entirely located proximate to the brain, i.e.
, below the skull, and is currently being pursued elsewhere [14
]. By contrast, our considerations have led to a “dual-panel” design with cortical and epicranial (subcutaneous) sections mounted on a common flexible polyimide substrate (Kapton – fabricated by MicroConnex inc.) as shown in . Distributing the microelectronics payload in this manner can have several advantages for future human applications.
The cortical “front-end” panel consists of the electrode array flip-chip bonded (using an epoxy dot technique described in [17
]) to a monolithic low power CMOS IC with preamplifier array and analog multiplexer and a single bypass capacitor. Placing only the preamplifier-multiplexer chip directly onto the cortical microelectrode array ensures that we capture the full neural signals from the brain with minimum environmental interference while minimizing direct heating of the brain tissue. (Our criterion for ΔT< 0.6 °C rise is conservative vs. typical thermal limit for medical implants).
A small flat ribbon cable, part of the common polymer substrate, with seven embedded gold wires for signal and power connects the cortical panel to the epicranial “back-end” panel. That panel, placed between the skull and the skin, holds the A/D converter and a digital control ASIC as well as the power and data transmission components. Most of the system power dissipation occurs in these components. Their placement can enable improved heat dissipation by surgically arranging “arterial plumbing”, and allows ready access for infrared data detection and for RF inductive coupling. It isolates the neural array from any magnetic forces that may be used to align the RF and telemetry units. Finally, it keeps open future options for telemetry and power supply.
shows an electronic block diagram of the present 16-channel system. Some of the key system requirements for the prototype units described below are: (i) mW level power consumption, (ii) ability to record both spikes and lower frequency field potentials, and (iii) a safe packaging and encapsulation strategy for eventual chronic use in humans.
Fig. 2 (a) Photographic images showing an implantable 16-channel microsystem with a dual-panel liquid crystal polymer substrate. A spiral pattern of RF power receiving coil is clearly visible in the backside image; (b) a block diagram of the dual-panel microsystem (more ...)
Front and back photographic images of a prototype microsystem are shown in , displaying components and interconnect wiring. (The cortical front-end shown contains several discrete components which have now been integrated into the preamplifier chip.) The “U-shape” for this construct was for testing in a specific non-human primate based on its anatomy. This was quantified by reconstructing the skull and the brain in a rapid-prototype plastic model from the monkey’s MRI and x-ray CAT scan images. The surface-mount components on the epicranial section are fastened to their contacts with Ag epoxy. The bare die of the digital controller IC is wire bonded to the gold wiring. The gold spiral on the back of the epicranial section is the RF inductive pickup coil. The largest components are the A/D converter (ADC) and the digital control ASIC. The analog-to-digital converter is an off-the-shelf, 12-bit, AD7594 (Analog Devices) packaged in a standard micro-SO8 package. The ADC operates at 34 to 40 Ksps per channel depending on the way power is supplied. This choice of a high sampling rate preserves all the subtle features of the spike waveforms so that useful information enabling distinction between single and multiunit activity recorded at a single electrode is available from such data.
The ADC receives a clock and a start-of-conversion signal from the digital controller IC that also supplies the channel address to the amplifier circuit. The controller is a custom integrated circuit built in the AMIS 0.5-micron process through MOSIS and has been described in more detail elsewhere [20
]. The controller has two other functions. First, it multiplexes the ADC data with a periodic synchronization code word that replaces the data from one channel. The external electronics that receive the neural data uses this unique code word to find the beginning of the serial data for the first channel. Second, the controller converts the multiplexed data into the drive current for a low-current, high-efficiency, vertical-cavity surface-emitting laser diode (VCSEL), which produces a peak optical output power of 2 mW for the optical telemetry. The VCSEL occupies less than 1 mm2
of substrate area. The controller derives its clock from either the RF inductive power loop or from modulation on the DC power source depending on the supply mode. The controller IC contains a comparator that regenerates the digital clock from a small sinusoidal signal separated from the appropriate source in the power module. Total system power consumption is approximately 12 mW in the present version including all parts of the implanted system.
The present integrated preamplifier-multiplexer chips has some excess noise at low frequencies that limits their utility for local field potential (LFP) measurements. They exhibit an equivalent amplifier input noise of 7.3 μVRMS
in a 50 Hz to 7.5 KHz bandwidth. Typical the 3 dB design bandwidth for each amplifier is approximately 7.5 KHz with a representative gain of 43 dB. We have carefully analyzed the noise sources [20
] and have developed a new design similar to that pioneered by Harrison and co-workers [21
] but have not yet used it in implantable units. This new design is anticipated to have an equivalent noise voltage of 4.7 μVRMS
and 45 dB of gain. (For design details see Ref. [20
].) The power in both current and anticipated designs is about 45 - 50 μW per channel depending on the exact power supply voltage. These performance values are viewed as acceptable for a practical chronic implant.
The entire microsystem of is presently encapsulated in polydimethylsiloxane (PDMS) for electrical isolation and mechanical flexibility. Surgical implant considerations require careful control of PDMS thickness to maintain flexibility in the tether and to prevent buildup over the electrode array. For images of the structure after encapsulation, see Refs. [12
]. The main functions of the encapsulation are to ensure (i) that electrical leakage current to the adjacent tissue is less than 10pA, and (ii) ionic leakage from the tissue to the electronic components is inhibited. For chronic implant applications, this presents a formidable challenge for all researchers in the field of implantable neural prosthetics. We view our initial approach, using PDMS (NuSil R-2188), as a useful starting pathway at least to subchronic or short-term (1 – 3 months) in-vivo animal testing.
To evaluate the performance of our ‘soft’ encapsulation we have soak tested samples for six months in saline at T= 52°C. The testing is done using a small test circuit board that is a simplified version of the complete implantable neurosensor. The test structure enables continuously monitoring the resistance between interdigitated conductors on the substrate surface as well as leakage current through the encapsulation material. The leakage currents are a proxy for the presence of ions that might have leaked through the encapsulation material. The test structure includes elements with all the same morphological characteristics that are encountered on the real devices and includes a working ADC. In a test of 10 sample devices, the leakage current between bath and circuit was found to typically vary between 1 and 10 pA at 3 VDC with no significant change over time. The ADCs provided appropriate data for the duration of the test. In spite of these results, it is clear that chronic implants will require a more reliably impermeable barrier. We are presently exploring combinations of soft organic polymers with inorganic thin film multilayer barriers or heterogeneous mixtures, solid solutions of inorganic molecules in polymers. Candidate inorganic materials include SiC, SiOx, or Si3N4.