Biomaterials 22 (2001) 1187}1193

The geometric design of micromachined silicon sieve electrodes in#uences functional nerve regeneration
Lars Wallman , Yanzhen Zhang , Thomas Laurell , Nils Danielsen *
Department of Electrical Measurements, Lund Institute of Technology, Lund University, P.O. Box 118, SE-221 00 Lund, Sweden Section for Neuroendocrine Cell Biology, Department of Physiological Sciences, Lund University, BMC F10, SE-221 84 Lund, Sweden Received 11 March 2000; accepted 22 September 2000

Abstract A neural interface could be used to control a limb prosthesis. Such an interface can be created by facilitating axonal regeneration through a sieve electrode and then register nerve signals intended to control the prosthesis. A key question is how to design the electrodes to ensure the best possible regeneration. Our previous studies have indicated that regeneration can be achieved using electrodes with square-shaped, 100;100 m, via holes (holes that axons will regenerate through). Other reports have indicated a suitable range of these holes between 40 and 65 m. In the present study we used silicon sieve electrodes with via holes of either 30 or 90 m. The transparency, i.e. the percentage of the total via hole area, of these electrodes was either 20 or 30%. The electrodes were inserted into a silicone chamber which was used to bridge a gap in a rat sciatic nerve. After 12 weeks of nerve regeneration electrodes with a hole size of 30 m and a 30% transparency had the most favourable result as judged by the regained gastrocnemius muscle force and the formation of reactive tissue inside the chamber. The sieve electrode transparency is crucial for ensuring regeneration. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Silicon; Nerve regeneration; Rat sciatic nerve; Sieve electrode

1. Introduction In order to develop a limb prosthetic device directly controlled by a physiological command a neural interface is needed. One way to create such a neural interface is to make a sieve electrode with a large number of via holes (i.e. the holes the axons will regenerate through) and insert the sieve electrode into a nerve regeneration chamber, usually made out of silicone [1]. The nerve regeneration chamber is then used to bridge a gap in a transected animal nerve and the regenerating axons will grow through the via holes and reinnervate the distal nerve stump. These sieve electrodes can be micromachined to contain a number of electrodes connected to the via holes in order to record action potentials from the axons that have regenerated through the device. If such a neural interface could be chronically implanted into a major nerve in a human being it could theoretically be used to register neural signals that would enable a more physiological control of a limb prosthesis. Several orig* Corresponding author. Tel.:#46-46-2220300; fax:#46-46-2223232. E-mail address: nils.danielsen@mphy.lu.se (N. Danielsen).

inal studies have demonstrated that peripheral axons have the capacity to regenerate through such sieve electrodes implanted in animal nerves and that it is possible to record action potentials from such electrodes [2}16]; for an extensive review on the subject of implantable bioelectronic interfaces in general see [17]. Usually the sieve electrode is made out of silicon [5] but also other materials such as polyimide [15] have been tried. A fundamental aspect when designing sieve electrodes is the geometric design of the electrode, i.e. the number of via holes, the hole density, and the size of each individual hole. Functional regeneration of the glossopharyngeal nerve has been achieved with a hole diameter of 2 m and a large number of holes (777) [6]. Other research groups plan to use larger square-shaped holes (64;64 m) in their active design [9] and this suggestion has gained support from another study, that indicated a suitable range of the via holes between 40 and 65 m [10]. The use of 40 m via holes gained further support in a study where the sieve electrodes were made out of polyimide [15]. When designing these devices one has to consider that a rat sciatic nerve contains about 27,000 nerve "bres; 6%

0142-9612/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 2 - 9 6 1 2 ( 0 0 ) 0 0 3 4 2 - 2

1188

L. Wallman et al. / Biomaterials 22 (2001) 1187}1193

are myelinated motor axons (1600), 23 and 48% are myelinated (6200) and unmyelinated sensory axons (13,000) respectively, and the remaining 23% are unmyelinated sympathetic axons [18]. The diameters of the various individual nerve "bre types must also be taken into account when designing these sieve electrodes. In mammalian peripheral nerves the largest myelinated nerve "bres (A ; that include proprioception and somatic motor) have diameters of 12}20 m [19]. If the ultimate goal is to make a device that could be chronically implanted into nerve stumps in an amputated limb to control a limb prosthesis, then it should be designed to at least facilitate motor nerve "bres, i.e. the via holes cannot be too small. In one of our previous studies, we reported successful regeneration through silicon sieve electrodes with via holes of 100 m, but not through 50 or 10 m via holes [14]. In this previous study we did not control for the so-called transparency factor, i.e. the percentage of the total via hole area in relation to the actual surface area of the electrodes (for the mathematical equation de"ning the transparency factor see Section 2). Therefore, the aim of the present study was to investigate the e!ect of the transparency factor on nerve regeneration through sieve electrodes. This was done by using sieve electrodes with two di!erent via hole size, 30 or 90 m, but the transparency of these sieve electrodes was controlled and constant regardless of the via hole size. The larger hole size, 90 m, was chosen since it was very similar (100 m) to the one we used in a previous study [14]. The smaller hole size (30 m) was chosen because it was smaller than the more generally accepted 40}65 m range [9,10].

Fig. 1. The process scheme for the fabrication of sieve electrodes. (1) A p-type 11 0 02 silicon wafer was single sided phosphorus doped with subsequent oxidation. The pattern for the perforated area was de"ned in the oxide layer on the doped side of the wafer. (2) The doped side was etched through the doped layer and the rear side was etched de"ning the size of the perforated membrane and the size of the chip. (3)The pn-etch stop system was activated and the etch automatically stopped at the doped layer leaving a 7 m thick perforated membrane (a) and via holes (b).

2. Materials and methods 2.1. Sieve electrode design and fabrication The sieve electrodes fabricated for this study were etched using the pn-etch stop technique, which is an electrochemical etch process, to precisely de"ne the thickness of the perforated membrane. The electrodes were micro-structured in 250 m thick p-doped 3 in 11 0 02 silicon wafers, that were single sided phosphorus doped with a subsequent oxidation step. The pattern for the via holes was de"ned on the phosphorus doped side of the wafer and on the other side the pattern for the membrane region and the supportive brim was de"ned by UV lithography. The wafer was etched in KOH solution (40 g KOH : 100 ml H O, 803C). When the via holes  had penetrated the doped layer, the pn-etch was activated (the n-doped layer was passivated) and the etch continued from the undoped side until the doped layer was reached and the etch automatically stopped. The membrane thickness was 7 m (Fig. 1) and the diameter

of the sieve electrodes was 3 mm. The thin perforated membrane made it possible to increase the transparency of the membrane. The required strength of the structure was achieved by the surrounding supportive brim. The fabricated sieve electrodes had via holes of two sizes, either 30 or 90 m (Fig. 2). These "gures re#ect the length of one side of the quadratic via holes, thereby giving individual hole areas of 900 m for the smaller hole size and 8100 m for the larger ones. Since the silicone chamber in the present study had an inner diameter of 1.8 mm the sieve electrode was designed in such a way that the via holes were evenly distributed over an area with a diameter of 1.8 mm. In order to achieve two di!erent transparencies (20 or 30%) in the sieve electrodes, the number of via holes was varied (Table 1). The transparency factor was de"ned according to the equation A N " & & , A 2 where T is the transparency factor, A the area of an & individual hole, N the number of holes and A the & 2 cross-section area of the silicone chamber. The silicone chamber was assembled according to the following scheme: Two 4 mm long silicone tubes with an inner diameter of 1.8 mm and an outer diameter of 3.2 mm were glued on each side of the silicon sieve electrode using silicone adhesive. Care was taken to avoid

L. Wallman et al. / Biomaterials 22 (2001) 1187}1193

1189

2.2. Surgical procedure The animals (female Wistar rats, 200 g) were anaesthetized with a 1.8 ml intraperitoneal (i.p.) injection of sodium pentobarbital (60 mg/ml) and saline in a 1:10 volume proportion. The sciatic nerve was exposed and transected at midthigh level proximal to the tibial}peroneal bifurcation. The proximal and distal nerve stumps were pulled 2 mm into each opening of the silicone chamber and secured to the chamber wall with single 9-0 perineurial sutures (Ethilon ) to ensure a gap of 4 mm between the two nerve stumps. The silicone chambers were "lled with sterile 0.9% saline before the surgical wound was closed in layers. All silicone chambers were implanted with the front side of the sieve electrode facing the proximal stump and the brim, rear side (Fig. 1), facing the distal. 2.3. Measurement of muscle contractility After 12 weeks of nerve regeneration the isometric contractility force of the gastrocnemius muscle was measured as previously described [14,20]. In short, the animals were reanaesthetized and the gastrocnemius muscle and the sciatic nerve were exposed. Supramaximal electrical stimuli were delivered to the sciatic nerve proximal to the silicone chamber by a Grass SD-9 stimulator at a frequency of 100 Hz for 0.6 ms. Meanwhile, the tetanic force of the muscle was measured with di!erent preloads of the muscle. The largest recorded values from both the injured side and the control side were saved and the measured contractility force on the experimental side was expressed as a percentage of the contractility force recorded on the contralateral control side. Ongoing studies using this evaluation method have indicated that there is a strong correlation between the regained contractility force and the number of regenerating nerve "bres [21]. After the recordings, the nerve specimens were excised and processed for light microscopy as described below. 2.4. Light microscopy After the muscle force recordings, the rats were killed and the silicone chamber with its content was excised and immersed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate bu!er, pH 7.2. A window was made in the chamber wall to allow the nerve structure maximal exposure to the "xative. Thereafter the chamber was cut away and a 2 mm long nerve piece was taken just distal to the sieve electrode. These nerve pieces were rinsed in bu!er and post-"xed in 1% osmium tetroxide and dehydrated in series of ethanol solutions. After dehydration the specimens were embedded in Agar 100 resin, and transverse sections 1.5 m thick were cut from the surface facing the sieve electrode from these 2 mm nerve pieces, stained with toluidine blue and examined by light microscopy. The total and endoneurial cross-sectional area of the nerve

Fig. 2. SEM of fabricated 30 or 90 m sieve electrodes. Rear ((a) 30 m * 20%) and front side ((b) 90 m * 20%) of the sieve electrodes processed according to the scheme depicted in Fig. 1.

Table 1 Regained muscle contractility force of the gastrocnemius muscle expressed as a percentage of the contralateral muscle for the four di!erent sieve electrode designs Type of sieve electrode Hole size ( m) * number of holes 30 30 90 90 * * * * 566 848 63 94 Transparency factor (%) 20 30 20 30 30$17 (n"6) 38$17 (n"7) 21$7 (n"8) 24$17 (n"7) Percentage of mean$SD muscle contractility force

The di!erences between the di!erent groups did not reach statistical signi"cance; n"number of experiments conducted.

getting any silicone adhesive inside the chamber and all chambers were inspected microscopically to ensure that no glue covered the perforated area. Before implantation in animals all chambers were sterilized overnight in 95% ethanol.

1190

L. Wallman et al. / Biomaterials 22 (2001) 1187}1193

Fig. 3. Light micrographs of the nerve structures distal to the sieve electrode. ((a) 30 m * 20%; (b) 30 m * 30%; (c) 90 m * 20%; (d) 90 m * 30%). The endoneurial area containing the majority of the myelinated axons was surrounded by a circumferential perineurial-like cell layer (a}d) outlined by arrowheads in (a). The nerve regenerates distal to the 90 m sieve electrodes (c, d) displayed a fascicular pattern corresponding to the via holes on the electrodes. This was more obvious for the 90 m * 20% sieve electrodes (c).

structure was calculated based on the microscopically measured total and endoneurial diameters. The endonurial area was de"ned as the area that contained the majority of the myelinated axons and that was surrounded by layers of circumferential perineurial-like cells [1]. The perineurial-like cell layer consists mainly of "broblasts and macrophages and can be regarded as a sign of reactive scar formation within the silicone chamber. The area of this perineurial-like cell layer can be calculated based on the "gures obtained for the total and endoneurial area by subtracting the endoneurial area from the total. Furthermore, the calculated area of the perineurial-like cell layers was expressed as a percentage of the total area in order to give an indication of the proportion of this reactive tissue layer. 2.5. Statistical methods For analysis of the whole material the Kruskal}Wallis test was used. For single comparison between di!erent transparencies but with the same via hole size the Mann}Whitney U-test was used; p(0.05 was considered signi"cant.

within the "rst hour after surgery because of respiratory problems. All of the remaining 32 rats survived for the planned 12 week regeneration period. 3.1. Measurement of muscle contractility Four rats had to be excluded from the measurements; three because of a collapsed sieve electrode (one rat from each of the groups: 30 m * 20%; 30 m * 30%; 90 m * 30%) and the remaining rat (30 m * 20%) because of technical problems with the muscle force measurements. These four rats were also excluded from further histological evaluation. Table 1 shows the measured contractility force expressed as a percentage of the contralateral side. Sieve electrodes with a hole size of 30 m and a transparency factor of 30% displayed the greatest regained relative muscle force (38%) and the "gure for 30 m * 20% sieve electrodes was 30%. For the sieve electrodes with 90 m via holes the regained contractility force was about 20%. 3.2. Light microscopy After 12 weeks of regeneration all sieve electrodes regardless of hole size and transparency factor had many myelinated nerve "bres distal to the electrode. The endoneurial area was surrounded by a thickened circumferential perineurial-like cell layer (Fig. 3a}d). The

3. Results A total number of 33 rats had sieve electrodes implanted. One of these rats (30 m * 20%) was killed

L. Wallman et al. / Biomaterials 22 (2001) 1187}1193

1191

Table 2 Measurements of total and endoneurial cross-sectional areas, and calculated cross-sectional area of the circumferential perineurial-like cell layer for the four di!erent sieve electrode designs Type of sieve electrode Hole size ( m) 30 30 90 90 Transparency factor (%) 20 30 20 30 1.17$0.26 1.06$0.52 1.12$0.33 1.17$0.38 0.55$0.22 0.66$0.30 0.59$0.18 0.69$0.22 0.62$0.25* 0.40$0.38 0.53$0.23 0.49$0.23 Total cross-sectional area (mm) Endoneurial cross-sectional area (mm) Area of perineurial-like cell layer (mm) Percentage of perineurial-like cell layer

52$15* 34$17 46$11 39$14

All values are mean$SD. The number of observations are the same as indicated in Table 1. The increased area of the perineurial-like cell layer and the corresponding increase in the percentage of this area for the 30 m}20% sieve electrode group as compared to the 30 m * 30% group were statistically signi"cant (indicated with *; Mann}Whitney U-test, p"0.0373).

regenerates from 90 m sieve electrodes had a di!erent morphological pattern as compared to the 30 m sieve electrode regenerates. The 90 m sieve electrodes resulted in a fascicular pattern corresponding to the via holes. This was more obvious for sieve electrodes with 20% transparencies (Fig. 3c) than for 90 m * 30% sieve electrodes (Fig. 3d). No signs of nerve "bre compression could be observed on the examined sections. Table 2 displays the results of the di!erent area measurements. The total area of the nerve regenerate for the four di!erent groups did not di!er. The di!erences in endoneurial area were also small. The calculated areas of the circumferential perineurial-like cell layers and the percentages of these areas showed greater di!erences between the four types of sieve electrodes. Sieve electrodes with 30 m via holes and the smaller transparency factor (20%) had the largest proportion (52%) of these circumferential cell layers and the 30 m * 30% the smallest (34%). This di!erence within the 30 m group was statistically signi"cant (Mann}Whitney U-test, p"0.0383).

4. Discussion The present study demonstrated that functional regeneration as judged by the regained muscle force can be achieved through micromachined silicon sieve electrodes with a via hole size of 30 m. It has been argued that silicon-based sieve electrodes present a signi"cant obstacle to nerve regeneration [15]. Previous silicone chamber experiments without inserted sieve electrodes have indicated that materials or substances such as gels placed between the two nerve stumps in the silicone chamber may a!ect the outcome of regeneration [1]. In the present study, we could achieve a 38% recovery of the gastrocnemius muscle force and this "gure should be compared to a 53% recovery of the gastrocnemius muscle force after nerve tubulization without an inserted

sieve electrode [20]. It is likely that any type of sieve electrode will a!ect nerve regeneration negatively and that one of the crucial factors to achieve regeneration through these electrodes is the transparency of the design. In our previous study we could not achieve acceptable functional regeneration through silicon sieve electrodes with 50 m via holes [14,22]. This was probably due to a much too small transparency factor, 9.6%. It is most likely that the geometric design of the sieve electrodes, i.e. the diameters of the via holes and the transparency of these electrodes, a!ects the functional outcome of regeneration through these devices. The most favourable result in the present study was achieved with 30 m via holes and the largest transparency factor (30%) as judged by the regained muscle contraction force and the smallest proportion of the circumferential perineurial-like cell layer. These electrodes also had the largest number of via holes (848; Table 1). That a large transparency factor is favourable is not surprising, the smaller amount of material blocking regeneration is expected to be advantageous. The question of which hole size should be used is more di$cult to answer and is still a subject for discussion. In the present study the electrodes with 30 m via holes gave the most favourable results. This hole diameter is in the range of that used (40 m) in a recent study using sieve electrodes made out of polyimide [15]. It has been argued that the most adequate via hole size is between 40 and 64 m [9,10]. One explanation for the less favourable results in the 90 m via hole group is that regenerating axons have to de#ect more for the 90 m group as compared to the 30 m group when regenerating nerve "bres from the proximal stump grow through the sieve electrode into the distal nerve stump (Fig. 4). A sieve electrode comprising large via holes will have fewer holes as compared to electrodes with smaller via holes if the transparency factor is kept constant (Table 1). Furthermore, in electrodes with large via holes, the holes are situated more apart as

1192

L. Wallman et al. / Biomaterials 22 (2001) 1187}1193

Fig. 4. The nerve has to de#ect more when regenerating through a chip with 90 m hole size (a) than for a sieve electrode with 30 m hole size (b) since the via holes are situated more apart.

compared to the smaller via hole design. This increased hole distance (Fig. 4), actually more material between the holes, may also serve as a physical block for the regenerating axons. Another advantage of using smaller via holes is that the average distance between the nerve "bre going through a hole and the recording electrode surface is shorter, thereby increasing the capacitive coupling between the "ring axon and the electrode which will give a better signal-to-noise ratio. The question of the via hole size is also related to what strategy is used to contact the regenerating axons and the overall purpose of the design. If the purpose is to contact and identify individual axons, the sieve electrode (number and size of the via holes) should match the average number and diameters of axons in the nerve as suggested by Navarro et al. [10] and the recording electrodes should contact individual holes [10,15]. On the other hand, if the purpose is to control a limb prosthesis, a more favourable strategy might be to ensure maximum regeneration by constructing sieve electrodes with large numbers of via holes and a via hole size of 30}40 m, thus giving a large transparency factor. These type of sieve electrodes can then be divided into a number of zones (e.g. 4, 6, or 8), where each zone is an electrode comprising a rather large number of via holes [16,22]. The recorded compound nerve signals can then be processed by an arti"cial neural network and used to control a limb prosthesis. It might not be necessary to identify individual axons but only to record distinctive patterns of nerve signals. This approach has the advantage that the control signal is not dependent on individual axons and their viability.

The choice of material for future sieve electrodes is important. Several factors have to be taken into consideration such as biocompatibility, electrical and manufacturing properties. Silicon seems to have acceptable biocompatibility properties. When silicon is tested against titanium in a rat soft tissue model [23] it elicited tissue reactions such as capsule formation and macrophage recruitment adjacent to the implant surface with the same magnitude as titanium [14,24]. However, the whole concept of introducing a transversely located sieve electrode in a silicone chamber between the proximal and distal nerve stump will of course elicit tissue reactions as indicated by the increased proportion of the circumferential perineurial-like cell layer. The present study indicates that this increase in the circumferential perineurial-like cell layer is more dependent on the transparency of the sieve electrode, since the proportion of this cell layer was reduced when the transparency was increased from 20 to 30% for both hole sizes (Table 2). Furthermore, the proportions of this perineurial-like cell layer observed for the sieve electrodes with 30% transparencies (34% for 30 m and 39% for 90 m; Table 2) are very similar to the "gure (39.3%) reported in nerve regeneration studies using ordinary silicone chambers without an inserted sieve electrode [25]. The disadvantage with silicon-based sieve electrodes is that it is di$cult to visualize the interface between the regenerated nerve "bres and the via holes. Scanning electron microscopy can give some information about the tissue structures in a via hole, but it is very di$cult to longitudinally cut the nerve structures going through the electrode due to the physical nature of the silicon. In this aspect, polyimide has an advantage since it is #exible and allows this type of sectioning. It has been reported that axonal compressive phenomena occur distal to silicon sieve electrodes [9,10] and that such phenomena do not occur when polyimide sieve electrodes are used [15]. The question is whether these axonal compressive phenomena are material dependent or depend more on the fact that the transparencies of previously tested siliconbased sieve electrodes have been too small. If the number of holes are too few, i.e. transparency factor too low, a large number of initially small regenerating axons will grow through the available holes. When regeneration proceeds and the nerve "bres mature, the diameter of these "bres will increase and can thereby theoretically become compressed. Even in the present study the available surface for the regenerating axons is only 30% of the inner diameter of the silicone tube. In fact, most of the available tube area is blocked by the sieve electrode. Previous ordinary silicone chamber experiments have shown that nerve regeneration is improved by simply increasing the chamber diameter [26]. It is our current working hypothesis that the transparency of the sieve electrode is one of the key factors in ensuring successful regeneration through these electrodes.

L. Wallman et al. / Biomaterials 22 (2001) 1187}1193

1193

Acknowledgements This study was supported by grants from NUTEK (Kofuma), the Swedish Medical Research Council (12712), the Swedish Foundation for Strategic Research (R98:007), and Bergvall's, Crafoord's, Tore Nilson's and Wiberg's Foundations.

References
[1] Danielsen N. Regeneration of the rat sciatic nerve in the silicone chamber model. Restor Neurol Neurosci 1990; 1:253}9. [2] Mannard A, Stein RB, Charles D. Regeneration electrode units: implants for recording from single peripheral nerve "bers in freely moving animals. Science 1974;183:547}9. [3] Loeb GE, Marks WB, Beatty PG. Analysis and microelectrode design of tubular electrode arrays intended for chronic, multiple single-unit recording from captured nerve "bres. Med Biol Engng Comput 1977;15:195}201. [4] Edell DJ. A peripheral nerve information transducer for amputees: long term multichannel recordings from rabbit peripheral nerves. IEEE Trans Biomed Engng 1986;33: 203}13. [5] Rosen JM, Grosser M, Hentz VR. Preliminary experiments in nerve regeneration through laser-drilled holes in silicon chips. Restor Neurol Neurosci 1990;2:89}102. [6] Bradley RM, Smoke RH, Akin T, Naja" K. Functional regeneration of glossopharyngeal nerve through micromachined sieve electrode arrays. Brain Res 1992;594:84}90. [7] Kovacs GTA, Storment CW, Rosen JM. Regeneration microelectrode array for peripheral nerve recording and stimulation. IEEE Trans Biomed Engng 1992;39:893}902. [8] Akin T, Naja" K, Smoke RH, Bradley RM. A micromachined silicon sieve electrode for nerve regeneration applications. IEEE Trans Biomed Engng 1994;41:305}13. [9] Kovacs GTA, Storment CW, Halks-Miller M, Belczynski Jr CR, Della Santina CC, Lewis ER, Maluf NI. Silicon-substrate microelectrode arrays for parallel recording of neural activity in peripheral and cranial nerves. IEEE Trans Biomed Engng 1994;41: 567}77. [10] Navarro X, Calvet S, ButmH M, Gomez N, Cabruja E, Garrido P, H Villa R, Valderrama E. Peripheral nerve regeneration through microelectrode arrays based on silicon technology. Restor Neurol Neurosci 1996;9:151}60. [11] Bradley R, Cao X, Akin T, Naja" K. Long term chronic recordings from peripheral sensory "bers using a sieve electrode array. J Neurosci Methods 1997;73:177}86.

[12] Dario P, Garzella P, Toro M, Micera S, Alavi M, Meyer J-U, Valderrama E, Sebastiani L, Ghelarducci B, Pastacaldi P. A microfabricated interface for neural recording and stimulation. J Micromech Microengng 1997;7:233}6. [13] Della Santina CC, Kovacs GTA, Lewis ER. Multi-unit recording from regenerated bullfrog eight nerve using implantable siliconsubstrate microelectrodes. J Neurosci Methods 1997;72:71}86. [14] Zhao Q, Drott J, Laurell T, Wallman T, Lindstrom K, Bjursten K LM, Lundborg G, Montelius M, Danielsen N. Rat sciatic nerve regeneration through a micromachined silicon chip. Biomaterials 1997;18:75}80. [15] Navarro X, Calvet S, RodrmH guez FJ, Stieglitz T, Blau C, ButmH M, Valderrama E, Meyer J-U. Stimulation and recording from regenerated peripheral nerves through polyimide sieve electrodes. J Peripheral Nerv Syst 1998;3:91}101. [16] Wallman L, Levinsson A, Schouenborg J, Holmberg H, Montelius L, Danielsen N, Laurell T. Perforated silicon nerve chip with doped registration electrodes; in vitro performance and in vivo operation. IEEE Trans Biomed Engng 1999;46:1065}73. [17] Heiduschka P, Thanos S. Implantable bioelectronic interfaces for lost nerve functions. Prog Neurobiol 1998;55:433}61. [18] Schmalbruch H. Fiber composition of the rat sciatic nerve. Anat Rec 1986;215:71}81. [19] Ganong WF. Review of medical physiology, 8th ed. Los Altos, CA: Lange Publications, 1977. p. 29. [20] Zhao Q, Dahlin LB, Kanje M, Lundborg G. Speci"city of muscle reinnervation following repair of the transected sciatic nerve: a comparative study of di!erent repair techniques in the rat. J Hand Surg 1992;17B:257}61. [21] Scherman P, Kanje M, Lundborg G, Dahlin LB. Continuous longitudinal sutures * a new and simple approach to repair nerve defects. VII. Congress of the Federation of the European Societies for Surgery of the Hand, Barcelona, Spain, June 21}24, 2000. [22] Wallman L. Silicon microstructured sieve electrodes for neural interfaces. Licentiate thesis, Department of Electrical Measurements, Lund Institute of Technology, Lund, Sweden. Report 2/99, ISSN 0346-6221, 1999. [23] Rosengren A, Danielsen N, Bjursten LM. In#ammatory reaction dependence on implant localization in rat soft tissue models. Biomaterials 1997;18:979}87. [24] Rosengren A, Wallman L, Bengtsson M, Laurell T, Danielsen N, Bjursten LM. Tissue reactions to porous silicon * a comparative biomaterial study. Proceedings from the Second International Conference on Porous Semiconductors * Science and Technology (PSST-2000), Madrid, Spain, March 12}17, 2000. [25] Danielsen N, Johansson BR, Dahlin LB. The e!ects of delayed nerve repair on nerve regeneration in a silicone chamber model. Restor Neurol Neurosci 1994;6:317}22. [26] Williams LR, Varon S. Modi"cation of "brin matrix formation in situ enhances nerve regeneration in silicone chambers. J Comp Neurol 1985;231:209}20.