A promising approach towards the creation of highly selective brain-machine-interfaces (BMI) is the utilization of flexible electrocorticography (ECoG) electrode arrays based on a polymer-metal-polymer-sandwich system . Unfortunately, these electrodes are very sensitive to mechanical loads and show failure due to breaking of the delicate metal structures. It has proven difficult to reach the desired flexibility of the electrode arrays without compromising the mechanical stability. In the past, mostly tensile loads were investigated for these kind of electrodes . To evaluate the stability under bending loads and to identify the unique failure mechanisms of such structures, an individually adaptable bending test procedure has to be developed which keeps tensile influences as small as possible.
2.1 Bending test principle
The test apparatus was designed according to the following simple bending principle. The flexible electrode arrays were mounted between two perpendicular plates and thereby forced to bend within the interjacent distance (Fig. 1). By changing the latter, the bending radius R could be varied and adapted to the test conditions. Enabling vertical movement for one of the plates allowed shifting of the bending’s vertex continuously over the electrode array. By choosing large movement distances, almost the whole array could be tested. As soon as the turning point was reached, the plate travelled back and the next cycle was started. These cyclic bending loads should simulate repetitive deformation of the electrode. The integrity of the electrode’s individual tracks was monitored via four wire resistance measurement. An important measure to describe the bending stability is the bending (or flexural) stiffness which is defined by a relationship between bending moment Mb, bending radius R, Young’s modulus E and moment of inertia I (Fig. 1 – b) and eq. 1).(1)
For these type of sandwiched polymer-metal electrodes, the bending stiffness is dominated by the metal layer which also defines the neutral plane within the compound. It is important to keep the asymmetric layer structure in mind when choosing a suitable bending radius.
2.2 Test apparatus
To enable an automatized bending and measurement procedure, a computer-controlled test setup was built. The following section is split in two sub-sections – the mechanical assembly and electronic control with data recording.
2.2.1 Mechanical assembly
According to the test principle described in section 2.1, the aim was to build two parallel planes where one had to be moveable in vertical direction. For that purpose, a linear guide unit with a gear drive belt (ZLW-0630, IGUS GmbH, Cologne, Germany) was vertically mounted on a base made of polymethylmethacrylate (PMMA). The aluminium guiding rail was equipped with a spacer which allowed fixation of the movable plate. The static counter plate was mounted perpendicular to the base on a horizontal slide (DryLin-T, IGUS), which allowed adjustment of the distance between the two plates and hence of the bending radius R for the test. Two different types of electrode chucks were developed, which could be stacked on the top end of each plate. For large electrode arrays with lateral dimensions exceeding 80 mm, a vacuum chuck was used. It possessed 17 holes, connectable to a vacuum supply, to reliably attach one side of the array. For smaller electrodes, such as the BMI ECoG electrode (cf. section 2.4), the second chuck type with its mechanical clamp was used. For the sake of completeness, both types are illustrated in Fig. 2 which depicts the mechanical parts of the assembly. To define the maximum test distance of the moveable plate, turning points had to be defined. For that purpose, sub-miniature switches (AH1FS, Panasonic, Hamburg, Germany) were mounted in custom sensor fixtures which could be placed in the linear unit’s side notch. As soon as the spacer of the vertical slide hit the switch, direction of movement was changed. For accurate placement of the sensor fixtures and a defined movement distance, a scale was attached to the back of the linear unit. According to the manufacturer, the linear unit needs a driving torque of 0.25 Nm. Two different engines (Type RB35, Modelcraft, Blaine, USA) were utilized to enable different movement speeds. Their main properties are summarized in Table 1.
2.2.2 Electronic control
Control of the electro motors as well as communication with a computer to initiate electrical measurement of the electrode tracks was achieved with a customized electronic circuit, subsequently referred to as motor control. A LabView (National Instruments, Austin, USA) program was written to enable automatic control of the up and down movement of the tester as well as data recording of the electrode track resistance development. After a defined amount of bending cycles, which could be set in the software, the movement was stopped and the electrode array was characterized by means of four-wire measurement. A multimeter (34401A, Agilent) performed the electrical measurement and a custom multiplexer (MUX) was used to toggle between the electrode channels, i.e. different tracks. A data acquisition microcontroller – DAQ (USB 6008, National Instruments) was utilized to control the MUX and the motor control circuit. A simplified block diagram to illustrate signal and control pathways is given in Fig. 3.
2.3 Design of test electrodes
To evaluate the effect of different track and contact geometries on the stability under bending load, test electrode arrays were designed and manufactured according to a well-established laser-based process. Details are described elsewhere . The conductive material for electrode tracks and contacts was MP35N (thickness 25 µm) embedded in medical grade silicone rubber (MED1000, Nusil, Carpinteria, USA). Four different test types were placed on the 80 x 80 mm2 electrode array (Fig. 4 –A, B, C, D). Each in two configurations, straight (1) and 180°-turn (2). For the test design below, the track width (70 µm) and contact size (1.1 mm) was scaled up by a factor of 1.5 (type A & C) and 2.0 (B & D) compared to the BMI ECoG described in earlier publications .
Electrode tracks were electrically connected to the multiplexer by means of MP35N wires (ř = 75 µm) which were resistance-welded to the bared track ends at the contact openings of the test array (Fig. 4 - top). The welds were stabilized and protected with a droplet of silicone rubber.
Three electrodes with structures as shown in Fig. 4 were tested to a limit of 35 000 bending cycles at a bending radius R = 6 mm. This radius originated from a worst case bending scenario during implantation. Motor 1 was utilized for this experiment (Table 1). Tracks which were still unharmed after passing this threshold were excluded from the data analysis and investigated separately. Within the first 100 cycles, the resistance was recorded every ten bendings. After that, an interval of 100 cycles was chosen for further recordings. A two-parameter Weibull failure mode analysis was performed for the dataset, to determine the characteristic lifetime T, where 63.2 % of the individual design types had failed. Details concerning the mathematical background can be found in . Furthermore the type of failure was visually determined.
2.4 BMI ECoG electrode array
A significantly smaller ECoG electrode array of a similar design as used for in vivo experiments  was characterized with the bending tester, too. Due to its lateral dimensions (ca. 20 x 36 mm2, exact dimension and layout can be found in ) it had to be mounted onto an overhead projector foil as carrier before testing. The bending radius was again set to 6 mm. The tracks’ integrity was measured after 10, 100, 1000 and 2000 cycles. The failure pattern was optically compared to the array which was implanted for 15 months in an animal (sheep) and had been potentially exposed to rough handling during surgery.
3.1 Test electrodes
In a first comparison of all tracks which had failed within the 35000 cycles, the Weibull distribution seemed to be an adequate fit for the analysis. The measurements did not reveal any difference between straight and 180°-turn tracks, hence both designs were summarized for the subsequent analyses. The characteristic lifetime T for wide tracks (scale factor 2) could be more than doubled to >25000 cycles compared to smaller tracks (scale factor 1.5). Furthermore, the Weibull shape factor β was increased from 1.21 (type A&C), which hypothesizes a random failure mode, to 2.28 (type B&D), indicating a wear-related failure due to the high amount of bending cycles (Fig. 5). A comparison of the individual designs (Table 2) showed a tendency for meander shaped small tracks (type A) to fail early due to e.g. manufacturing issues (β<1). Regarding the characteristic lifetime, the analysis did not result in a significant difference between meander and straight track design. The significantly increased shape factor for wide tracks without meander (type B) indicated a Gaussian rather than a Weibull distributed failure probability function.
A high standard error for and T could be explained by the sample size for the individual designs under test (N=8-12). The most prominent failure throughout all designs was breakage at the transition, track to contact pad (Fig. 6). It is important to mention that multiple failures of a single track were included in this evaluation, e.g. each contact pad could break at the top and bottom transition. The sample size in Fig. 6 is hence a measure for all conceivable failure spots per design type on all three arrays tested.
3.2 BMI grid electrode array
The BMI grid array possessed very delicate structures (width = 70 µm) which were significantly more sensitive to mechanical loads compared to the test tracks in section 3.1. All but seven tracks were damaged after 2000 bending cycles (79 % failure rate). In 22 cases, the transition, track to pad failed (Fig. 7 - right), which was again the most apparent weak spot. In comparison to the similar array which was utilized in vivo for 15 months  and obviously deformed (Fig. 7 - center) the track - pad transition failed in seven cases. Even though the kind of bending might have been different and less frequent during handling, surgery etc., the failure mode was comparable (Fig. 7 – left).
The introduced bending test procedure for flexible electrode arrays allows to reliably predict the stability of different electrode track geometries. Failure modes were also comparable to previous in vivo experiments. Based on these findings, design rules can be set to create electrodes of superior reliability for future cortical implants.
Gierthmuehlen, M.; Wang, X.; Gkogkidis, A.; et al., “Mapping of sheep sensory cortex with a novel microelectrocorticography grid”, The J. comp. neurol., 2014; 522 (16): 3590–3608 Google Scholar
Schuettler, M.; Stiess, S.; King, B. V.; Suaning, G. J., “Fabrication of implantable microelectrode arrays by laser cutting of silicone rubber and platinum foil”, JNE, 2005; 2 (1): S121-8 Google Scholar
About the article
Published Online: 2015-09-12
Published in Print: 2015-09-01
Conflict of interest: MS is CTO and FK and TS are consultants of Cortec GmbH. Material and Methods: Informed consent: Informed consent is not applicable. Ethical approval: The research related to animals use has been complied with all the relevant national regulations and institutional policies for the care and use of animals.