Neural implants are technical systems that are mainly used to stimulate parts and structures of the nervous system with the aid of implanted electrical circuitry or record the electrical activity of nerve cells. Their application in clinical practice has given rise to the fields known as “neuromodulation” and “neuroprosthetics” (or neural prostheses). From the experience gained by the early experiments in the 1960s, miniaturization technologies, material sciences and the progress in medical and especially neuroscientific knowledge evolved and paved the way to these novel applications in therapies of neurological diseases and rehabilitation of lost functions in clinical practice.1, 2 Neuromodulation, namely the stimulation of central nervous system structures to modulate nerve excitability and the release of neurotransmitters,3 alleviates the effects of many neurological diseases. Deep brain stimulation helps patients suffering from Parkinson's disease to suppress tremor and movement disorders. It is also a treatment option for severe psychiatric diseases like depression and obsessive-compulsive disorder. Vagal nerve stimulation has been applied first to treat epilepsy3, 4 but has now expanded to psychiatric diseases and many more applications are under development in preclinical and clinical trials. The most commonly implanted device in the neuromodulation sector is the spinal cord stimulator, used to alleviate chronic pain and to treat incontinence.5 More than 130,000 patients have benefitted from these implants5 that derive from cardiac pacemakers, first developed decades ago.
Neural prostheses aim to restore lost functions of the body, either sensory, motor or vegetative. An early example can be dated back to about 1970 when Giles Brindley implanted the first electrodes around the sacral nerves of spinal cord injured persons to manage their bladder function.6, 7 Other implants have been developed in parallel to help patients suffering from stroke or from spinal cord injury. Motor implants to restore grasping,8 stance and gait9, 10 as well as ventilation11 by electrical stimulation of the diaphragm have been developed and introduced into preclinical studies or even as commercial products to the market. However, the number of patients that benefit from these systems is relatively low, in part due to some technical shortcomings, but mainly as a result of the limited performance of the implants in patients due to their individual course of injury. In combination with a limited market, it is economically quite unattractive for companies to develop and approve a new device, since the reimbursement is uncertain and the sale volume is (too) low. Sensory implants to restore hearing, so-called “cochlear implants,” are one of the main success stories of neural prostheses. More than 150,000 patients have been implanted with these technical systems, that stimulate the nerve cells in the inner ear at several sites when the sensory cells (hair cells) are no longer present, for example, due to aging, diseases (meningitis or Meniere's disease) or by certain drug treatments.12, 13 Congenitally deaf children, as well as adults who have lost their hearing at a later point in life14 have been implanted and were able to hear and to communicate via the telephone with these implants. Recently developed implants can access the brain stem15 and midbrain auditory structures16 when tumors have destroyed the pathways from the ear to the cortex, to restore at least some sound perception. In these cases, there is still room for a lot of improvement since the speech processors are “borrowed” from cochlear implants and are not yet optimized for the neuronal target structures.
The latest technological progress in miniaturization technologies has enabled the development of retinal prostheses to restore vision through implantation of complex electrical stimulators into the eyes of blind people.1, 17 Clinical trials have proven the feasibility of the approach, but there are still many limitations to overcome and thus it is likely that commercial products will not be available within the immediate future.
The concept of controlling technical devices and neural prostheses by “thoughts” currently drives research in the field of brain-machine interfaces, where a large variety of different materials and approaches compete to become the first reliable solution for a clinical application.18 Unfortunately, enthusiasm about the technological opportunities masks the risk and side effects that come along with implantation. Therefore, benefits and detriments have to be carefully considered in any medical and surgical treatment, and ultimately the patient should give the final consent for implantation to occur.
All neural implants have to fulfill general requirements to become approved as a medical device: They must not harm the body and should stay stable and functional over a certain life-time which is in most cases in the range of decades. Hermetic packages made of ceramics or titanium are state of the art2 to protect the implant electronics from moisture and ions. These packages are implanted in most cases in a place that is quite far away from the neuronal target tissue to prevent any undesired interaction or damage. The key challenge for any neural implant is the proper design of the neuro-technical interface. Multiple electrical contact sites have to get in close contact with neural tissue to selectively stimulate subsets of nerve cells. Nerves are delicate and structures of soft tissue get easily damaged by hard materials especially when forces due to movements occur. Polymers have been found the optimal material class when requirements of little response to implantation, long-term stability in a hostile environment, low material stiffness (i.e., high material flexibility), and good electrical insulation of metallic conductors have to be combined in a single material.
In this review, we have compiled the information that seemed to us most valuable to enter the field of neural prostheses and to enable the reader to make up his own mind about neural implants. First, we summarize the fundamental requirements of a material-tissue interface with the nervous system. Next, the most common polymers and their material properties are summed up, since the most widespread materials for substrate and insulation materials of electrodes and cables are polymers. Following this, the use of polymers and their performance in different neuro-technical interfaces with the peripheral and central nervous system are presented for the various designs and applications. Finally, we summarize the topic of polymers for encapsulation and packaging of implantable electronics; however, a full discussion goes beyond the scope of the current review.
FUNDAMENTAL SPECIFICATIONS OF NEURO-TECHNICAL INTERFACES
Even though millions of people worldwide benefit from artificial joints and cardiac pacemakers, it does not mean that the challenges to implant a technical material in the human body are completely solved or even fully understood. In addition to their own special requirements, all implants have certain fundamental specifications in common to be stable and functional. These specifications will be explained using the example of neuro-technical interfaces. The interface of neural implants forms the boundary of technical system and delicate soft tissue, that is nerve cells that are surrounded by supporting cells like glial cells in the central nervous system or fibrous tissue such as the perineurium and the epineurium in the peripheral nervous system. The mechanical properties of the biological tissue have to be taken seriously into account when selecting materials for implant manufacture as well as the anatomical constraints and conditions of the implantation site considering space and movement of the nerve versus muscles, skin and bones. From the engineering point of view, proper and detailed target specifications including the details and limitations of the intended use are mandatory to make a proper design that combines a sufficiently high robustness with the necessary complexity of the technical system and not the highest possible one.
Any implant is designed to have a minimal impact on the body. However, any surgical intervention is accompanied by an inflammatory response as a normal physiological response to this intervention. In the presence of an implant, this response tends to be increased and extended depending on the chemical composition of the implants surface. A nonspecific foreign body reaction is initiated19 that -colloquially spoken- tries to eat up the implant or to wall it out. Specific immune responses with antibody mediated immune responses hardly occur and should be prevented by the material selection but also by the cleanness (i.e., the amount of germs and dirt on the implant). The aspect of surface biocompatibility deals with all viewpoints of chemical and biological interaction of an implant with the surrounding tissue. This biological process chain starts in any case with an unspecific protein adsorption (also called biofouling) that triggers the foreign body response. In the best case, it ends up with a defined and stable encapsulation of fibrotic tissue without harming the implant, a so-called bioinert reaction. This encapsulation might be beneficial with respect to the spatial fixation of the implant. However, since we deal with implants to record electrical signals and to electrically stimulate nerve cells, this electrically insulating encapsulation always results in an increase of current or voltage threshold for stimulation and in a decrease in the signal-to-noise ratios during recording until a steady state is reached. Therefore, substrate materials and coatings must be chosen that the reaction after implantation is minimized and that reactive cells are transferred into their inactivated state after the healing reaction is terminated.20 If these specifications are met, the material can be considered a reasonable biomaterial.21 One prerequisite for such a material is that it must not cause large inflammation after surgical intervention, and cell behavior must not be altered by toxic products that diffuse out of the material itself. These basic material investigations are identified in in vitro cytotoxicity tests with standardized cell cultures. The international standard ISO 10993 “Biological evaluation of medical devices” describes test systems, procedures and evaluation schemes to classify an implant as biocompatible or not. Cytotoxicity testing helps to reduce animal experiments and allows assessment of different materials due to standardized and application specific cell lines. Alterations in cell morphology and metabolism are good indicators for toxic chemical groups or elutes and the surface energy of the devices under test. Polymer materials have proven (see “Polymer Materials” section) their suitability as implant materials. Their surface properties can be even improved by chemical and bio-chemical surface modifications. In this review, however, we will focus on the “inherent” material properties without any additional modification.
Surface biocompatibility is necessary but depicts only one aspect of biocompatibility. Structural biocompatibility refers to mechanical interaction between the implant and the surrounding tissue and includes weight, shape and flexibility (Young's modulus). For a long time, mechanical mismatch has been assigned to cell and tissue damage and the following release of mediators as a result of the implantation event initiating the inflammatory cascade. Therefore, a lot of research has been conducted to reduce insertion damage especially in the central nervous system (for a review, see ref.22) but also in the peripheral nervous system to prevent collateral damage of the nervous tissue by movements of the implant. The mismatch of mechanical properties of the technical material and target tissue leads to cellular reactions that attack and eventually encapsulate the implant23 and results in a less effective electrical performance as already described in “Surface Biocompatibility” section. Recently,22 it has been shown in central nervous system implants that micromovements due to mechanical mismatch also lead to a chronic inflammation that results in glial scars around electrode carriers and finally electrode failure. These investigations on how the organism reacts to a certain material, shape and texture have to be performed in chronic in vivo experiments in animals before any device is allowed to enter a clinical trial according to the ISO 10993 and other directives, for example, the medical device directive and the active implantable device directive (AIMD) in the EU. Polymers with their large variability in stiffness and their ability to establish multilayer substrates with gradients in Young's moduli are excellent candidates with respect to structural biocompatibility in neural implant applications.
Most of the chronic implants stay within the human body for decades. Even if the battery powered electronics have to be exchanged every 5 to 10 years, the neuro-technical interface, that is the electrode, usually remains implanted. The technical term “biostability” summarizes different chemical aspects with respect to material stability and system integrity.23 Metals should not corrode, and substrate and insulation layers should not delaminate or degrade. In polymers, hydrolytic, oxidative and enzymatic degradation may occur and can be accelerated by pH changes and voltages on integrated interconnect lines. In vitro soaking tests, for example, in physiologic saline, Ringer's solution or cell culture media allow a first approximation of the biostability of the materials and are often performed at higher temperatures to accelerate the diffusion processes and thereby their influence on aging and the mean time to failure. General aspects on testing procedures are also described in the ISO 10993. Care has to be taken in the models of accelerated aging to predict the mean time to failure. Failure mechanisms in polymers do not always follow diffusion processes but depend on temperature initiated processes in some cases that are not described by the Arrhenius equations that are commonly used in the field of implant manufacturing.24 However, the results must be validated by in vivo tests for the most promising material candidates to exclude additional enzyme or foreign body reactions that could not have been foreseen in in vitro models. Again, the ISO 10993 guides the applicant through the necessary experiments that have to be done before a medical device approval can be passed.
Changes in Implant Performance due to the Intended Use
Neural implants should not interfere with the mechanical, chemical and physiological properties of nerves (e.g., trauma caused by surgical intervention) as already discussed earlier. In “active medical devices,” the official regulatory term for any recording and stimulation interface to the nervous system, the presence of electronic systems and their use must not alter nerve behavior either. Design measures have to be taken to prevent injection of input bias currents and the generation of DC voltages during recording and also in the “off state” of stimulation devices, for example, by decoupling capacitors and high quality electronic devices. Electrical stimulation might stress and harm the surrounding tissue if certain charge limits are exceeded. Corrosion products of the electrodes, pH shifts during stimulation and the decomposition of water and proteins have to be prevented25 to ensure implant integrity as well as to prevent nerve damage. Risk management to identify risks and hazards and to counteract them with adequate measures and modifications (in Europe according to ISO 14971) is a mandatory step in the development of any medical device. Despite these precautions, implanted electrodes become encapsulated with an electrical insulating fibrous tissue capsule26 that deteriorates the recording of nerve signals and increases the excitation thresholds in electrical stimulation. Additional electrode coatings with drug loaded polymers or polymer based conductors22 are under development but have not been proven long term stability yet.27
Reliable performance of an implant is based on neuro-technical interfaces with a reproducible input–output behavior over the lifetime of the implant after the acute healing phase. So far, no implantable technical sensor or actuator remains really “stable” over years; specifically, they have to be retuned or recalibrated after a certain timeframe for example, due to sensor drift or the accommodation to a changing environment. Adaptive systems would be needed to follow the target signal during learning or to switch to a more reliable recording site if inconsistencies are present over a longer time period. In addition, the response characteristics of biological sensors and actuators as well as psychophysical aspects have to be taken into account if biological signals from the muscle or nerve are taken as input command or control signal. Most biological sensors show a phasic-tonic behavior, that is they reduce their temporal output even though the environmental input remains constant. These nonlinearities have to be taken into account as well as fatigue, training dependent signal amplitudes and limited access to state variables for control purposes. Recent developments in control paradigms and signal processing approaches include neuroplasticity28 during task execution to improve control in neural prostheses and brain-computer interfaces. These aspects, however, go far beyond the scope of this review.
Polymer materials have been established as excellent materials in chemistry, automotives and electronics to interconnect different components, to electrically insulate conductors and to survive harsh corrosive environments. Food industry as well as pharmaceutical industry use polymers to package beverages, food and drugs with different materials and compounds. Therefore it is quite natural that the medical device industry has focused its attention to polymers for medical devices in general and especially for encapsulation and insulation of active implants. Since no polymer is hermetic in respect to water vapor,29 target specifications concern the flexibility and stiffness compared to (nervous) tissue, electrical insulation but also handling properties and manufacturing technologies to obtain the desired feature sizes, thicknesses and mechanical performance, which is often a compromise between strength to withstand the implantation process and the flexibility to prevent damage of the target tissue. This section presents those polymer materials that have been already established in many clinical applications or in research developments and serve as carrier or as insulation material for neural implants.
Polyimides are a branch of commercial available polymers most widely used in various aspects of microelectronics and also, within the last 30 years, in biomedical applications. Generally used as an insulation or passivation layer, polyimides provide protection for underlying circuitry and metals from effects such as moisture absorption, corrosion, ion transport, and physical damage. Furthermore, it acts as an effective absorber for alpha particles that can be emitted by ceramics, and as a mechanical stress buffer.30 Key properties are: thermoxidative stability, high mechanical strength, high modulus, excellent insulating properties, and superior chemical resistance (for values see Table 1). Typically, polyimides are available as photo-definable and nonphoto-definable versions whereas photo-definable polyimides tend to have a higher moisture uptake that limits their use in vivo; the latter will therefore not be discussed further.37 Synthesis of polyimides is achieved by adding a dianhydride and a diamine into a dipolar aprotic solvent (like N,N-dimethylacetamide or N-methylpyrrolidinone) which rapidly forms poly (amic acid) at room temperatures. This precursor of polyimide can be easily stored, shipped or used to form thin films, coatings and fibers. Conversion of poly(amic acid)s to the designated polyimides is most commonly performed by thermal imidization. For wafer level manufacturing, this involves spin-coating of the precursor onto the wafer, which specifies the thickness of the layer, a prebake at modest temperature (∼120 °C) to drive the solvent partly out of the layer, which makes it more sticky, and a curing step at high temperature (∼350 °C) in nitrogen atmosphere.38 Metal can be deposited afterwards onto the polyimide by various means, for example, vapor deposition or sputtering, and encapsulated by a second layer of polyimide. Using reactive ion etching (RIE) with oxygen, the polyimide-metal-polyimide stack can be patterned and electrodes and/or interconnection sites opened. The resulting devices can be peeled off the wafer using tweezers. Overall, processing polyimides is similar to conventional microelectronic processes, yielding low production costs, high pattern accuracy and high repeatability.
Table 1. List of the Electrical, Mechanical, and Thermal Properties of Chosen Polymers
Although polyimide, especially the BPDA/PPD type [see Fig. 1(b)] which is most often used as biomaterial and commercially available under the trademark of DuPont's PI2611 or UBE's U-Varnish-S, is not certified according to the aforementioned ISO 10993, various groups have proven its biocompatibility, low cytotoxicity and low hemolytic capacity, both for bulk materials39 and long-term implanted electrodes.40 Existing applications are manifold and include peripheral nervous system (PNS) and central nervous system (CNS) implementations, such as cuff and intrafascicular electrodes or shaft and ECoG electrodes, respectively (see “Classification of Neural Interfaces” and “PNS Interfaces” section). Devices made of polyimide have elicited only mild foreign body reactions in several applications in the peripheral and central nervous system showing good surface and structural biocompatibility.41–44 They have proven to be biostable and functional for months in chronic in vitro and in vivo studies.24, 45
Since Kipping in 1904 assigned the name “silicone” to the group of synthetic polymers whose backbone is made of repeating silicon to oxygen bonds and methyl groups, this material and its applications have flourished. It is probably the most widely used material among the synthetic polymers for biomedical applications today. Later, a more specific nomenclature was developed and the basic repeating unit became known as siloxane and the most common silicone is polydimethylsiloxane or PDMS [see Fig. 1(c)]. Since the methyl groups can be substituted by a variety of other groups, for example, phenyl, vinyl or trifluoropropyl, enabling the linkage of organic groups to an inorganic backbone, silicones can be prepared with combinations of unique properties. They are used for example, as insulators in electronics, as moulds in semiconductor manufacture, as sealants or adhesives in the construction industry, as well as in numerous pharmaceutical and medical device applications.46 Their key-features for use in biomedical applications include physiological indifference, excellent resistance to biodegradation and ageing, and high biocompatibility (for values see Table 1). A further significant property is the high permeability to gases and vapors that is about 10-fold when compared to natural rubber, while acting as ion barriers. A previous biodurability study showed no changes in the material properties after 2 years of implantation in test and control specimens, and no evidence of biodegradation could be detected.47 Furthermore, implants utilizing silicone encapsulation such as the Brindley bladder stimulator have already been in clinical use since the 1970s and proved to be stable over a period of about 25 years in vivo, after which the silicone rubber was reported to become more brittle.7, 48 Silicones can be processed either by spin-coating, resulting in thinner film thicknesses, or by molding techniques, which enable their use in a variety of applications. In biomedicine, PDMS is usually used as encapsulation and/or as substrate material. When encapsulating a device for use in vivo, special attention should be paid to a number of aspects such as the adhesion of silicone to bulk material and void free deposition and curing of the silicone rubber, since these will significantly contribute to osmotic reactions occurring when silicone is immersed into ionized water (e.g., the body environment).49–52 As substrate material, PDMS is often spin-coated to achieve a defined and uniform layer. In a next step, a patterned metal foil is placed onto the uncured silicone rubber and a second PDMS layer is spin-coated on top. After curing, the polymer-metal-polymer stack can be patterned by laser ablation, wet or dry etching; all techniques have their specific pros and cons.53–55 Photo-definable PDMS is also available but not in implantable grades.
Common biomedical applications include cardiac pacemakers, cuff and book electrodes in the PNS, and cochlear implants, bladder and pain controllers, and planar electrode arrays in the CNS. Many of these systems are commercially available and found in common clinical settings (for an overview see e.g., ref.2). PDMS is one of the most successful polymers since it results in only mild foreign body reactions (as cured material in active implants), is extremely stable and keeps its flexibility. It is an excellent insulator and has clinical approval according to USP class VI for unrestricted use in chronic implants. Process technology is well established for various manufacturing technologies.
Parylene is the common name for polyparaxylylene (PPX), a group of linear, noncross-linked and semicrystalline polymers, which belong to the thermoplasts. Since the discovery of the manufacturing process in the mid 20th century, more and more parylene types have been developed, that differ only slightly in their properties. Parylene layers are deposited in a vapor deposition polymerization56 process, using the dimer of the adequate parylene type as starting substance. This dimer is heated up until it vaporizes and later on splits into a monomeric gas. When the gas reaches the deposition chamber, it cools down and polymerizes on the target. This deposition process allows a conformal coating of the target from all sides and even sharp edges and crevices under components are covered.57–59 Typical layer thicknesses reach from a few hundred nanometers until several micrometers, depending on the coating machine. The deposited layers are compatible to MEMS processing and can be structured by reactive ion etching.
Parylene C [poly(dichloro-p-xylylene), Fig. 1(a)] is the most popular parylene type for the use in biomedical applications, due to the well suited combination of electrical and barrier properties. It is used as substrate60–62 or encapsulation63–67 material for many kinds of biomedical microdevices. Its good biocompatibility68 (FDA approved, USP class VI), chemically and biologically inertness, good barrier properties, slippery surface and its functionality as an electrical insulator predestines parylene C for the use as substrate or encapsulation material for implanted neural prostheses.
In recent years another parylene type called parylene HT arose, which has similar properties, but can withstand higher temperatures. The first electrode arrays using parylene HT as substrate material, have already been produced.69
Parylene C has been established as one of the encapsulation materials for chronic implants due to the aforementioned properties and due to its approval as material for unrestricted use in implants. However, due to our own experience, the handling properties of thin sheets of parylene C are not as good as those of polyimide in comparable thickness. The material is more fragile and is not as strong and robust, for example, in substrate integrated microelectrode arrays. Its advantage, however, is the deposition technology at room temperature that does not interfere with connection and assembling technologies.
Liquid crystal polymers (LCPs) represent a separate material class among the polymers. They are built up of rigid and flexible monomers which are linked to each other, and hence they can organize in aligned molecule chains with a crystal-like spatial regularity. The main properties of LCPs are high mechanical strength at high temperatures, extreme chemical resistance, low moisture absorption and permeability, and good barrier properties for other gases.
Originally, LCP was used as a high-performance thermoplastic material for high-density printed circuit boards fabrication and semiconductor packaging. Today, many different types of LCPs are available, including LCPs which are specially designed for use in medical engineering (FDA approved, USP class VI). Commercial LCP material is supplied in sheets with predefined thickness from 25 μm to 3 mm. These sheets are melt-processible and can be structured by laser machining and reactive ion etching. LCPs are not yet widely-used in biomedical applications, but some first approaches to use LCPs for the fabrication of flexible electrode arrays have been developed.70
LCPs were promised to become the new shooting star in neural interfaces due to the low water uptake and the manufacturing technology. However, the enthusiasm of the first scientific presentations has not been transferred to many groups. Results from chronic in vivo studies have to show first, if the promised performance can be achieved and if the results are better than with already established materials.
SU-8 is a multicomponent photoresist, based on epoxy SU-8 resin including a photo-acid generator (PAG) compound and incorporated solvent. It has interesting properties, which make it a very attractive material for a wide range of applications including micromachining, micro-optics, microfluidics, and packaging. SU-8 is highly transparent for wavelengths >400 nm and is chemically and mechanically stable (see Table 1).71 The photoresist is commonly exposed with conventional UV radiation, although i-line (365 nm) is the recommended wavelength. Upon exposure, cross-linking proceeds in two steps: first, formation of a strong acid during the exposure step, followed by a second, an acid-catalyzed, thermally driven epoxy cross-linking during the postexposure bake.34, 35 The oligomer is depicted in Figure 1(d), the eight functional groups allow for high degrees of cross-linking after photo-activation.72 An additional hard bake can further cross-link the imaged SU-8 structures and therefore improve the sidewall smoothness and mechanical stability.73
Biocompatibility tests were performed by different groups using a baseline battery of ISO 10993 physiochemical and biocompatibility tests and found minimal irritation after one- and 12-week implantation periods in rabbit muscles, as well as after a 54 week implantation in rats. Cytotoxicity showed a reactivity less than grade 2 (mild reactivity) and no steam or gamma sterilization-induced damage was observed.74, 75 However, due to the complex nature of biocompatibility, it cannot be concluded that SU-8 meets all specifications necessary to meet the full ISO 10993 requirements for implants.
In biomedical applications, SU-8 is commonly used as a substrate material, for example, in shaft electrodes for CNS interfaces due to its improved flexibility compared to silicon,76 as a guiding structure for regenerating axons in sieve electrodes,77 as a wave-guiding core in electrodes for optical stimulation,78 and as a microfluidic channel in multimodal electrodes that have the capacity to deliver for example, pharmaceuticals next to electrical stimulation or recording.79
SU-8 might be an alternative material to silicon. It is biocompatible and the processing costs are cheaper compared to the relatively expensive silicon micromachining. However, since no microelectronic circuitry can be integrated, it has to prove superior performance at the material-tissue interface to become a real competitor to already established silicon shaft electrodes.
Classification of Neural Interfaces
Neural interfaces and implants are used for many applications in the human body and can be manufactured from many different materials. Although there are a huge variety of types of implants and interfaces, they can be divided up into various categories, with some similarities and differences between them. One immediate feature that many have in common is design of cables of packages for electronics and batteries. In this section, we will focus on the neuro-technical interface itself. One system for classifying implants, introduced earlier in this review, is whether the implants are used to target the central nervous system (CNS) or peripheral nervous system (PNS). In general, approaches for CNS and PNS are different since brain structures need different access methods than nerve bundles in the periphery. However, there is a general tendency in both:43 greater spatial selectivity requires a higher level of invasiveness (Fig. 2). The question to select an adequate interface always should be: Which selectivity do I need for the intended use of my target specification and which degree of invasivity is adequate? In other words: Is the benefit of the implant large enough that I can justify the risk of possible damage due to the implantation? Detailed solutions to interface technical devices with the CNS and PNS at different levels of invasivity will be displayed in the following sections.
Since we like to focus on materials that are “flexible” instead of “stiff,” to better match the mechanical properties of nervous tissue, we do not describe wire or silicon based neuro-technical interfaces (for an overview, see e.g., ref.80) but focus on polymer materials. These polymer materials have been selected in a way that they adapt to the shape of the neuronal target tissue as well in the PNS as in the CNS and follow motions of the tissue in the micro as well as in the macro scale. They are the enabling materials to manufacture electrode arrays, either in small or large scale. All of the applications that are presented below have shown that the interfaces are only little reactive due to material and shape, that is only mild foreign body reactions could be observed and are stable over the implantation periods. A lot of care has been taken in all designs and developments to find the “optimal” shape of the implantable nerve interfaces to reduce the risk of device failure due to tethering forces as well as the damage of the target tissue due to “clumsy” designs. However, this expert knowledge cannot be transferred in simple design rules but is a long process of continuous exchange and communication with the “end user” in experimental neuroscience and clinical research and applications.
In the following, a brief overview of interfaces to contact with the peripheral nervous system is given. To stay consistent with Figure 2, the different interfaces will be explained in order of increasing invasiveness, from low to high. Since a large variety of approaches emerged over the last few decades, it is not the intention of the authors to provide a complete list of interfaces, but rather to present the key approaches and concepts underlying electrode design and manufacture. For more details see references 81–84.
Extraneural cuff electrodes are commonly made out of PDMS [Fig. 3(a)] or polyimide, and encircle the nerve completely.86, 87 Hence, the invasiveness is limited to the preparation of the nerve, which itself stays untouched. Cuff electrodes contain a number of electrode sites on the inner surface facing the nerve and have been investigated over decades and have finally been transferred into clinical applications.88, 89 Despite their advantages of simplicity of handling and the ability to stimulate and record general activity from the outer parts of the nerve, they still have a number of limitations. Firstly, their selectivity is limited to subgroups and superficial fibers in the nerve. Secondly, the nerve can be damaged due to micromotion of the electrode array, especially in peripheral nerves of the limbs.83 To achieve a better selectivity Tyler and Durand designed a variation of the cuff electrode that slowly penetrated the epineurium, loose connective tissue that is placed between and around nerve bundles, without compromising the perineurium. Thus, the electrode sites are placed within the nerve trunk but outside the nerve fascicles. The slowly penetrating interfascicular nerve electrode (SPINE) achieves this aim through blunt elements extending radially into the lumen of a PDMS tube that encloses the nerve.90 However, after histological evaluation it was shown that the shape of the nerve was actually deformed from an elliptical shape into a flatter ribbon like shape giving access to deeper fascicles, that is nerve bundles that are surrounded by the perineurium of the nerve. From these results, a new electrode design was extracted, the flat interface nerve electrode (FINE), which reshapes the nerve into a more electrically favorable geometry [Fig. 3(b)]. Since this reshaping requires the slow application of a relatively high force, only moderate flattening of the nerve is possible without inducing nerve damage.91, 92 Recently, Schiefer et al. implanted a multicontact FINE in the femoral nerve of humans and showed high selectivity in restoring knee extension and hip flexion by functional electrical stimulation at least in an acute experiment.93
Penetrating electrodes comprise the next level of invasiveness. Intrafascicular electrodes are placed within a distinct fascicle in the nerve and have direct contact to the targeted fibers and are, hence, more invasive than extraneural electrodes. The placement of the contact sites increases the signal-to-noise ratio (SNR) of recordings and enhances stimulation and recording selectivity. Stimulation of targeted fascicles can be achieved with little cross-talk to adjacent fascicles while complete recruitment of the nerve fascicle is possible with low stimulation intensities. The longitudinal intrafascicular electrode (LIFE) and the transversal intrafascicular multichannel electrode (TIME) are the most recent examples [Fig. 3(c,d)]. Both designs implement polyimide substrates and platinum metal tracks and active sites. As the names suggest, LIFEs are implanted longitudinally within individual nerve fascicles whereas TIMEs are implanted transversally through the designated nerve and fascicles.44, 94 Since TIMEs penetrate the whole nerve and, thus, contact more fascicles on its way, they are expected to have a higher selectivity than LIFEs but comparative studies are not available yet. Rossini et al. implanted LIFEs for 4 weeks in the median and ulnar nerves of humans and reported reproducible and localized hand/finger sensation while stimulating and stable selective recordings.95
If the nerve is already severed, for example, after an amputation trauma, regenerative electrodes can be applied. Sieve like electrodes made of polyimide incorporate multiple via-holes and electrodes through which regenerating axons are coaxed to grow through. These electrodes have shown good long term in vivo stabilities and a decent regeneration of axons.40, 96 Recent approaches facilitate a three-dimensional electrode that has a guidance structure, made out of SU-8 or polyimide, through which regenerating axons should grow. These guidance structures are in principal channels that can be filled with nerve growth factor or other bioactive solutions and can even incorporate electric active sites which enable additional recordings.97, 98 Although a high degree of selectivity can be achieved, sieve electrodes require the transection and regeneration of nerves, and thus can only be ethically applied in already transected nerves. Moreover, time is needed for the regenerating axons to grow through the structure, thus, precluding acute experiments.
Summarizing the results of different PNS nerve interfaces, we can conclude that PDMS with embedded metal tracks and electrodes is superior when low integration density and medium size implants have to be developed that need to be robust in first place. Microsystem technologies using polyimide as substrate material is only advantageous and beneficial if extremely small structures and large scale integration is required. In these cases the approach is mandatory and justifies the longer development time and the higher development and manufacturing costs.
Neural interfaces to the central nervous system have a variety of applications. Depending on function and implantation site, implant requirements and properties are different, and thus many different electrode designs exist. However, two general concepts developed over the last decades: precision mechanics implants with metal sites and wires encapsulated in PDMS and micromachined approaches with silicon as bulk material to manufacture multiple shafts in a single device that look like a brush or a nail bed. Depending on implantation site and application these implants have different shapes and numbers of electrode sites. The most common known and well established neural implant is the cochlear implant (CI). These systems are commercial available (MED-EL, Vienna, Austria; Cochlear, Lane Cove, Australia; Advanced Bionics, Valencia, California; Neuroelec, Vallauris, France) and are now implanted for about 30 years. The neural interface of such a system is a multichannel electrode [Fig. 4(a)] that is inserted into the scala tympani of the cochlea. Due to the special shape and the fragile structure of the cochlea, the electrodes have to be flexible, and hence they are made of soft materials such as silicone rubber99–101 or polyimide.102, 103 Today's research mainly focuses on achieving a higher selectivity and hence better hearing quality.13, 104, 105
Another well known application for interfacing with the CNS, despite still being in the research and development phase, is the restoration of vision. Experiments in the late 1960s and early 70s demonstrated that blind humans can perceive electrically elicited phosphenes in response to ocular stimulation, with a contact lens as a stimulating electrode,106 several groups worldwide work on the development of either epiretinal prostheses17, 69, 107–114 with implantation of the device into the vitreous cavity on the retinal surface or subretinal prostheses115–120 with implantation of the prosthesis in the potential space between the neurosensory retina and the retinal pigment epithelium. It is obvious that for a retina implant no stiff and bulky substrate or housing materials can be used, due to the limited space and the shape of the implantation site. The use of polymers for the stimulating electrode arrays is more or less obligatory. Generally they are made of silicone,108 polyimide,17 parylene C110 or a combination of them. Figure 4(b) shows a parylene-based microelectrode array with 1024 stimulating sites (60 of them connected), developed by Rodger et al.,69 that was chronically implanted for 6 month onto the retina of canines. The preceding array with 16 stimulation sites, embedded into silicone rubber, was already implanted up to 18 month into three patients with retinitis pigmentosa,111 which were able to perform simple visual tasks better than before implantation.
Interfacing the optical nerve is the most invasive approach to restore sight. There, silicone cuff electrodes are used,121–123 which are similar to those, which are used to interface the PNS (see PNS section). But due to the high risk during surgery and the poor resolution, interfacing the optical nerve is not the first choice to restore vision. The visual cortex is also used as interface for vision prostheses either using polymer coated wires or silicon microneedles as implants (for a review, see ref.1). Since we like to focus on flexible polymer based neural implants, we do not describe this exciting research here.
Another site of the CNS that is possible to interface is the spinal cord. In the 1980s, Brindley introduced a sacral anterior root stimulator for bladder control in paraplegia, with book electrodes [Fig. 4(c)] as neural interface.124 These electrodes entrap the sacral roots and are made of silicone rubber. Until 1994, 500 patients received this implant system.125 Silicone rubber electrodes are used until today to interface the spinal cord, but researchers also developed electrodes, using other polymer materials like parylene as substrate material.69 However, clinicians have learned to handle the silicone based electrodes in the spinal canal and implants have shown excellent performance with respect to long-term stability. Therefore, evolutionary developments seem to have a higher success rate to get transferred into clinical practice than new revolutionary designs in which stability, performance and side effects are not clear.
A much wider field of applications offers the direct contacting of brain tissue. To interface the brain, two completely different electrode designs are possible. The first and less invasive possibility is the use of epicortical electrode arrays. These are two-dimensional electrode arrays which can be placed epidurally or subdurally on the cortex. They have to adapt to the anatomical structure of the cortex, making polymers necessary as substrate material. The first arrays were made of silicone rubber and were used for stimulating the visual cortex126 or recording for localizing epileptogenic foci.127 Today silicone rubber arrays are commercial available (e.g., Ad-Tech Medical Instrument Corporation, Racine, WI) for the use in clinical practice to locate the seizure focus during the presurgical diagnosis of epilepsy.128 To obtain a higher spatial resolution, researchers try to scale down the electrode diameters and pitches and to scale up the amount of recording/stimulating sites. This is feasible with MEMS technology using polyimide as substrate material.129–134 Figure 4(d) shows a 252-channel polyimide-based electrode array, which was implanted onto the visual cortex of a monkey,134 and was after 4.5 month still able to record signals.
In all epicortical electrode arrays the flexibility of the material and the interface is the most important property. Mechanical strength of the implant is important only during the implantation procedure. Afterwards, supporting substrate structures could be dissolved, for example bioresorbable silk fibroin, as recently proposed by Kim et al.135
Generally, epicortical electrodes record local field potentials, but depending on the application, a higher selectivity is required. This is possible with a more invasive approach, more precisely the use of penetrating electrodes, which are directly inserted into the cortex. To be able to penetrate the brain tissue, a certain stiffness of the electrode shaft is required. Therefore, iridium wires, tungsten wires or silicon shafts were used over a long time period and polymers like parylene C or polyimide served as insulating material.65, 136–139 But stiff silicon shafts cannot adapt to the mechanical conditions of the brain tissue and, hence, micromotion of the brain, due to breathing and the heart beat, constantly injures the brain tissue. For this reason, researchers are currently trying to develop penetrating electrodes using more flexible materials like polyimide [Fig. 4(e)], parylene C, SU-8 or benzocyclobutene (BCB).60–62, 139–147 However, the use of flexible substrate materials is accompanied by further challenges. It is difficult to insert flexible probes into the cortex and hence several approaches to overcome this problem were developed, like bending the electrode shafts,148 coating or filling the shafts with degradable/dissolvable materials,61 partly attaching a silicon backbone layer,142, 143 or using an insertion shuttle149 to make the shafts stiff enough for insertion and flexible enough to decrease the tissue damage.
The discussion with its accompanying hypotheses how stiff electrodes should be for intracortical implantation to elicit minimum tissue reaction in chronic implantation is still not finished. Implants need either certain stiffness or adequate tools to get inserted into the brain. Coating might be beneficial but should be either stable over time or resorbable without eliciting additional adverse reactions. Nowadays, there are not yet enough data acquired to take the final decision which path has to be followed to get the “best” intracortical interface. More research has to be done before these intracortical microstructures are mature enough to be transferred into clinical applications.
From Interfaces Towards Active Implants
Neural implants interconnect a neuro-technical interface (the electrode array) with electronics and energy supply. Reliable implants with market approval for chronic implantation either place the electronics and battery in a hermetic package or consist of hermetically sealed components that are encapsulated in a nonhermetic coating.150 Even though neural implants have different designs on the interface part, the concepts and paradigms for hermetic packages are very similar. Most packages consist of metal (titanium) or ceramic (alumina) packages with a very limited number of hermetic “feedthroughs” that interconnect the electronic circuitry inside the package with the electrodes and sometimes a coil for energy supply and data transmission outside the package. Polymer materials are widely used as final encapsulation and material-tissue interface but they are not applicable as hermetic encapsulation since all polymers are NOT hermetic according to the definitions in international standards. The established and proven technologies are sufficient for the established neural implant applications. Nevertheless, more sophisticated feedthroughs and packaging techniques together with reliable hybrid assembly techniques have to be developed to accept the challenges of high channel implants for vision prostheses or brain-machine interfaces with anticipated 1000 channels and the biological space restrictions in the eye and the brain. Furthermore, much time, effort and financial resources are necessary to transfer a device from the actual proof of concept to an approved and certified biomedical product, which is accepted by physicians and patients alike. In the case of the cochlear implant, there was a period of ∼15 years between the first clinical trial and the acceptance according to NIH guidelines.151
Polymers have been the enabling material to develop and produce the vast majority of neural-technical interfaces that exist today. The ability of certain polymers to adapt to the conditions of the surrounding tissue is crucial to manufacture stable interfaces that can work reliably over many years without harming the body. The success stories of cochlear implants, bladder management systems or epicortical electrode arrays for epilepsy management, which are all commercial available systems and are already used in clinical practice, show that polymers can fulfill these requirements. PDMS is the most successful material and proved to combine only little tissue reactivity with excellent long term stability and reliability. In combination with precision mechanics for packages, cables and electrode sites, neural implants have predicted lifetimes close to the life expectancy of humans. Unfortunately, the technology has reached its limit of complexity and further miniaturization and integration density is hard to overcome with existing concepts and philosophies.
Fabrication technologies for polymers to produce interfaces with high counts of electrodes within limited space (and thus high selectivity) are available. The materials, however, have never been in a chronic implantation in a large number of patients, so far. The main challenge is still to avoid the hydrolytic, oxidative and enzymatic degradation due to the harsh environment of the human body or at least to slow it down to a minimum which enables the interface to work over a long time period, before it finally has to be exchanged. Therefore, it is mandatory to carefully select the appropriate materials, keeping in mind that each application has its own specific requirements. Materials like polyimide show promising results but no company is on the market that is currently willing to deliver materials with the certificates that are recommended from the national authorities for its use in medical devices. Without this approval according to national or international standards, the most “adequate” materials from a technical and practical standpoint are precluded due to legal and economic reasons. In addition, micromachining technologies that allow large scale integration of electrodes often use much thinner layers as clinically established implants. The stability of polymer insulation materials as well as on electrical interconnects and electrodes -aspects that have been intentionally neglected in this review- have to fulfill higher standards to survive the same implantation time; degradation as well as corrosion rates must be orders of magnitudes lower than in precision mechanics implants since we start with material thicknesses in the nanometer and micrometer range. Nevertheless, since the list of appropriate polymers is constantly increasing and possibilities to manufacture tailor-made surfaces that further enhance the materials' behavior become more suitable, it seems feasible that long lasting polymer interfaces beyond PDMS will be available in the near future.
Part of the work has been funded by the German Federal Ministry of Education and Research (BMBF) in the Bernstein Focus Neurotechnology Freiburg/Tübingen: “The Hybrid Brain” (grant no. 01GQ0830) and by the European Union in the 7th Framework Program (grant CP-FP-INFSO 224012/TIME) for the TIME project (Transverse, Intrafascicular Multichannel Electrode system for induction of sensation and treatment of phantom limb pain in amputees). The authors thank Ben Townsend and Martin Schuettler for their constructive criticism regarding this manuscript.
Christina Hassler received the Dipl.-Ing. in microsystems engineering in 2008 from the University of Freiburg, Germany. Later she joined the group for Biomedical Microtechnology at the Faculty of Engineering (IMTEK), University of Freiburg as a Ph.D. student. Her interest in research focuses on polyimide-/parylene-based intracortical microelectrodes with biodegradable coatings.
Tim Boretius received the Dipl.-Ing. in microsystems engineering in 2008 from the University of Freiburg, Germany. Later he joined the group for Biomedical Microtechnology at the Faculty of Engineering (IMTEK), University of Freiburg as Ph.D. student. His interest in research focuses on polyimide based microelectrodes, active coatings and the assemblage of the same.
Thomas Stieglitz received the Dipl.-Ing. in electrical engineering in 1993 (TH Karlsruhe), the Dr.-Ing. in 1998 and qualified as a university lecturer (Habilitation) in 2002 (both from the University of Saarland, Saarbruecken, Germany). From 1993 to 2004 he was with the Fraunhofer-Institute for Biomedical Engineering, St. Ingbert/Germany. Since October 2004, he is a full-time Professor for Biomedical Microtechnology at the Faculty of Engineering (IMTEK), University of Freiburg. His research interests include biocompatible assembling and packaging, microimplants, and neural prostheses.