**Implant Treatment in Atrophic Maxilla by Titanium Hybrid-Plates: A Finite Element Study to Evaluate the Biomechanical Behavior of Plates**

#### **María Prados-Privado 1,2,\*, Henri Diederich <sup>3</sup> and Juan Carlos Prados-Frutos <sup>4</sup>**


Received: 31 May 2018; Accepted: 24 July 2018; Published: 25 July 2018

**Abstract:** A severely atrophied maxilla presents serious limitations for rehabilitation with osseointegrated implants. This study evaluated the biomechanical and long-term behavior of titanium hybrid-plates in atrophic maxilla rehabilitation with finite elements and probabilistic methodology. A three-dimensional finite element model based on a real clinical case was built to simulate an entirely edentulous maxilla with four plates. Each plate was deformed to become accustomed to the maxilla's curvature. An axial force of 100 N was applied in the area where the prosthesis was adjusted in each plate. The von Mises stresses were obtained on the plates and principal stresses on maxilla. The difference in stress between the right and left HENGG-1 plates was 3%, while between the two HENGG-2 plates it was 2%, where HENGG means Highly Efficient No Graft Gear. A mean maximum value of 80 MPa in the plates' region was obtained, which is a lower value than bone resorption stress. A probability cumulative function was computed. Mean fatigue life was 1,819,235 cycles. According to the results of this study, it was possible to conclude that this technique based on titanium hybrid-plates can be considered a viable alternative for atrophic maxilla rehabilitation, although more studies are necessary to corroborate the clinical results.

**Keywords:** atrophic maxilla; titanium hybrid-plates; finite element analysis; biomechanical analysis

#### **1. Introduction**

The reconstruction of an atrophic maxilla has always been a challenge [1] because of anatomical and clinical factors due to the serious limitations for conventional implant placement [2]. These limitations are related to the amount of bone, which remains insufficient for the conventional placement of a dental implant [3]. The maxillary bone volume has been classified, among other authors, by Cawood and Howel in five grades (I to V). Grades IV and V are considered as extreme atrophies [4]. The most common alternatives in atrophic maxilla rehabilitation are bone grafting [5], pterygoid [6] or zygomatic implants [7], bone regeneration (with or without mesh) [8,9], and finally, short implants [10].

Bone grafting is the most common technique in the reconstruction of an atrophic maxilla [2]. The goal of hard tissue augmentation is to provide an adequate bone volume for ideal implant placement and to support soft tissue for optimal esthetics and function. Zygomatic implants present a viable alternative for this kind of treatment given their design with self-tapping screw and the appropriate length as this kind of implant can be placed in the bone with very good quality and excellent mechanical behavior [2,7]. Pterygoid implants have the advantage of allowing anchorage

in the pterygomaxillary region, eliminating the need for sinus lifts or bone grafts. Additionally, pterygoid implants can eliminate posterior cantilever and improve axial loading [2,11]. Finally, short implants are widely used and have demonstrated their efficiency on implant treatment in atrophic jaw and maxilla [12,13].

The first hybrid plates were introduced by G. Scortecci in July 2000 and were first used in the same year in a patient with a fractured atrophic mandible. They had a large base plate (25, 33, or 43 mm long, 7, 9, or 12 mm wide) [14]. These kinds of plates can be adjusted to the maxilla curvature and put in the best place to maintain occlusal function [15].

There are technical differences between the hybrid plates employed and the sub-periosteal implants. The first difference is that the hybrid plates are made of titanium grade II and machined from a block, and the sub-periosteal implants are made of chrome cobalt and cast individually. The second difference is that hybrid plates are flexible and can be adjusted in situ while sub-periosteal implants are rigid and modeled using an initial impression of the bone site. Sub-periosteal implants can only be cemented to the prosthesis while hybrid plates have a screw connection. To use them, it is necessary to make a groove in the bone to receive the plates so that they may become osseointegrated, which are then fixed with osteosynthesis screws.

Hybrid plates are made of titanium grade II, which is the main cp Ti used for industrial dental implant applications [16]. It is recognized that titanium and its alloys are biomaterials with the best in vivo behavior [17]. Due to the excellent biocompatibility of this material, it is very common to use it for biomedical applications such as dental implants ad hybrid-plates [18]. This biocompatibility is provided by the following properties of titanium: low level of conductivity, high corrosion resistance, thermodynamic state at physiological pH values, and low ion-formation tendency in aqueous environments [19,20].

In that sense, the main differences between hybrid plates and other alternatives to treat atrophic maxilla as pterygoid, zygomatic, or short implants are a titanium alloy. Implants are mostly manufactured by titanium alloy Ti6Al4V.

The protocol employed in this study is called Cortically Fixed @ Once (CF@O). It is an alternative to conventional implant placement for severe atrophied maxilla and mandible. This technique uses plates and pterygoid implants. Plates are fixed to the bone with osteosynthesis screws. There are four types of plates available that differ in size and morphology, although in this study, only two of them were used. This technique has its origins in basal implantology, which was developed by Dr. Scortecci in the early 1980s when he proposed the Diskimplant®, a disc implant system that was inserted laterally, and which he refined over the next few years [21]. Several basal implants were developed during the 1980s and the 2000s with different geometric forms and with perforations over the surface to improve the blood supply around the implant [14].

In the last few decades, the finite element method (FEM) has become very popular in the field of biomechanics as it is a useful tool to numerically calculate aspects such as stresses and strains, and to evaluate the mechanical behavior of biomaterials and human tissues, considering the difficulty in making such an assessment in vivo [22,23].

Dental implants and their components including hybrid-plates are subjected to cyclic loads. Therefore, fatigue of materials is introduced in all dental rehabilitations. Results with a good accuracy are essential in dental studies as fatigue is very sensitive to many parameters [22].

Owing to the variety of techniques for atrophic maxillary implant rehabilitation, a new technique based on innovative hybrid titanium plates is described and numerically analyzed. The aim of this study was to evaluate the biomechanical and long-term behavior of CF@O plates on a completely edentulous and atrophic maxilla by employing a finite element analysis and a probabilistic fatigue approach.

#### **2. Materials and Methods**

#### *2.1. Description of the Protocol*

The CF@O protocol is an alternative to the existing treatment of the atrophied maxilla and mandible. This technique is less invasive than the conventional procedures, and implants can be loaded with a definitive restoration after 6–10 days. Between the surgery and the definitive prosthesis, the patient has a provisional prosthesis, which is a fixed immediate loading prosthesis made of resin installed on plates.

This protocol does not need a sinus lift for the rehabilitation in cases of atrophied maxilla nor bone graft in maxilla and mandible and is based on both traditional implantology methods, combined with the most modern tools.

The Cortically Fixed @ Once protocol is applicable to edentulous maxillae and mandibles, and to unilateral and bilateral edentulism in maxilla and mandible. This protocol is indicated in the following assumptions [24]:


This protocol can be used in patients aged between 35 and 90 years and is performed under local anesthesia. It can be done in nearly all cases of atrophy except in the case of egg shell everywhere in the maxilla and 10 mm of residential bone in the mandible. A stereolithographic model based on CT scan is made. After surgery, a temporary fixed bridge is employed. Amoxicillin is prescribed (2 g a day for 10 days) and in case of pain, ibuprofen 600 (1–3 g a day). Finally, there are no restrictions on food after one month after surgery.

The treatment plan starts with signed patient consent. Then, an open flap in the maxilla, as in this study, is made from the left tuberosity along the crest till the canine region. Two hybrid plates HENGG-1 and HENGG-2 are fixed with osteosynthesis screws and covered. The flap is then closed on the left and right with polytetrafluoroethylene polymer (PTFE) monofilament nonabsorbable suture.

Depending on the atrophy, three types of prosthesis are available: a metal acrylic for a big atrophy and a metal ceramic or zirconia when there is less atrophy with enough bone. Another point to consider is the contribution of the pterygoid implants, which contribute to prosthesis fixation.

From 2013, 155 patients between the ages of 38 and 85 (95 were female and 60 were male and 105 were in the maxilla and 45 in the mandible) were treated with the protocol detailed previously, resulting in three failures of plates in the maxilla and two in the jaw. These lost treatments were related to infection processes in the soft tissues. After a follow-up period of one year, there was a clinical and radiographical check to confirm that the plates were fixed and without complication.

The finite element analysis employed in this study simulated a real case of a 58-year-old female, with patient consistent, who wanted fixed teeth in the maxilla in a compromised bone. The treatment consisted of two pterygoid implants, four hybrid plates fixed with osteosynthesis screws, and a metal acrylic bridge ten days later [25]. In this instance, only the plates were analyzed.

#### *2.2. Plates*

The plates used in the CF@O protocol are very thin, lightweight, and highly flexible, and therefore may be adapted to any bone anatomy. In this study, the two plates employed are detailed in Figure 1:

**Figure 1.** Plates employed in this study: (**a**) HENGG-1; (**b**) HENGG-2.

The HENGG-1 plate is appropriated for atrophied maxilla and is fixed with the zygomatic bone and the palate. The HENGG-2 plate is recommended for premaxilla, and the retromolar region in case of pencil mandible.

The plates are milled in a single piece and may be tilted in two axes to ensure that the implant fits the bone perfectly by manual shaping, making them isoelastic and able to mimic bone. They are minimally invasive and totally adjustable; they can be tilted at 90 degrees and the number of vents needed can be reduced as required, depending on the bone available at the site. They can also be twisted to fit the mandibular anatomy. They are stabilized and fixed by osteosynthesis screws, which give a strong cortical anchorage.

#### *2.3. Finite Element Reconstruction*

Geometric characteristics of the plates employed in the present study are shown in Figures 2 and 3.

**Figure 2.** Plate HENGG-1.

**Figure 3.** Plate HENGG-2.

All three-dimensional plates were adjusted to the anatomic characteristics of the maxilla (Figure 4b) as the common procedure in a real case (Figure 4a).

**Figure 4.** Plate adjusted to the anatomic characteristics of the maxilla: (**a**) Plate deformed before being placed in the patient; (**b**) three-dimensional model of a plate deformed.

The finite element model reproduced the case detailed previously, which is represented in Figure 5. Geometry of the maxilla was obtained using CT and transformed to the STL format. Slice increment was 0.5 mm, according to other studies in the literature. All data in DICOM format were imported into the software package Mimics 10.0 (Materialize, Leuven, Belgium) for the construction of the 3-D model. Plate HENGG-1 was placed in the molar region and HENGG-2 in the premaxilla.

Finally, Figure 6 represents the three-dimensional finite element assembly employed to reproduce the clinical case detailed in this study.

The maxilla in STL format was imported into SolidWorks 2016 (Dassault Systèmes, SolidWorks Corp., Concord, MA, USA) where the assembly with the four plates was done (Figure 6). The trabecular bone was 1 mm thick [26].

**Figure 5.** Model employed to reproduce in the finite element model.

**Figure 6.** Finite element model employed in this study.

#### *2.4. Material Properties*

All materials were considered isotropic, linearly elastic, and homogeneous. The properties of the materials are detailed in Table 1.


**Table 1.** Material properties employed in this model.

#### *2.5. Meshing*

Mesh generation was done in SolidWorks 2016 (Dassault Systèmes, SolidWorks Corp., Concord, MA, USA). All components were meshed with a fine mesh and all regions of stress concentration that

were of interest were manually refined. The three-dimensional model presented a total of 432,404 nodes and 294,104 elements. The convergence criterion was a change of less than 5% in the von Mises stress in the model [28] (Figure 7).

**Figure 7.** Finite element mesh (isometric view).

#### *2.6. Boundary Conditions and Loading Configuration*

The model was subjected to a rigid fixation restriction in the upper and lateral maxilla to prevent displacement in the *x*, *y*, and *z* axes (Figure 8). Plates were in contact with the maxilla and a nonpenetration condition was also added to prevent interferences during the execution process between the plates and the maxilla.

A load of 100 N [29,30] was directly applied perpendicular to the area where the prosthesis was fixed to the plate as detailed in Figure 8.

**Figure 8.** Three-dimensional posterior view of maxilla with load application and boundary conditions. Loads are represented in blue arrows and the rigid fixation restriction is represented in green.

#### *2.7. Probabilistic Fatigue Model*

In addition to the previous deterministic finite element analysis, a probabilistic fatigue model at the crack nucleation stage was also implemented. This stage is the most important regarding dental components and, hybrid plates, life, in particular [31]. As Riahi et al. detailed in their study, the probabilistic finite element method is a viable tool to estimate the influence of the stochastic properties of loads, material properties, and geometry on the response [32]. The methodology employed in this study was based on a cumulative damage B-model, which is constructed from the Probabilistic Finite Element Method (PFEM) results and computed for every random variable here considered: the Young's modulus (105 ± 10 GPa) and the applied loads (100 ± 10 N) [31]. The input random variables considered in this study were handled via its first order Taylor series expansion. Once all the sensitivities of the random variables are known, it is possible to apply the mean and variance operator. All the sensitivities of the random fields involved, such as displacements field, strain field, and stress field can be obtained.

Bogdanoff and Kozin (B-K) created a number of probabilistic models of cumulative damage based on ideas from Markov chains. This study employed one they called the B-model of unit steps, for its simplicity and suitability to the physical description of the process of fatigue in the crack initiation stage. The hypotheses that serve as a basis for the expansion of the B-K unit step model are [33]:


As damage cycles have been defined as constant severity, the Probability Transition Matrix (P) will be unique and expresses the probability that each DC must be in the same level or the probability will jump to the next DC. This matrix depends on the *pj* (probability of remaining in the same DC) and *qj* (probability of jumping to the next DC) and is detailed in Equation (1).

$$P = \begin{pmatrix} p\_1 & q\_1 & 0 & \dots & 0 & 0 \\ 0 & p\_2 & q\_2 & 0 & \dots & 0 \\ 0 & 0 & p\_3 & q\_3 & \dots & 0 \\ \vdots & \vdots & \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & 0 & \dots & p\_{b-1} & q\_{b-1} \\ 0 & 0 & 0 & \dots & 0 & 1 \end{pmatrix} \tag{1}$$

The new vector *px* is a vector showing the distribution of damage levels for time *t* = *x*. Using the results of Markov chains, vector *px* is:

$$p\_{\mathbf{x}} = p\_{\mathbf{x}-1}P = p\_{\mathbf{0}}P^{\mathbf{x}} \text{ with } \mathbf{x} = 0, 1, 2, \dots \tag{2}$$

Finally, to compute the fatigue life estimators, Neuber's rule and a random formulation of the Coffin and Basquin–Manson expressions were employed. Neuber's rule relates the levels of elastic stress and strain obtained by a linear elastic analysis with actual levels of stress and strain, in accordance to the elastic-plastic behavior material [34].

Coffin, for the elastic component of deformations, and Basquin and Manson, for the elastic-plastic component, proposed a nonexplicit relationship between the fatigue life cycles in the nucleation stage of a component and the amplitudes of strain. This relation is shown in Equation (3)

$$\frac{\Delta \varepsilon\_{\varepsilon p}}{2} = \frac{\sigma\_f^{\prime}}{E} \left(2N\_f\right)^b + \varepsilon\_f^{\prime} \left(2N\_f\right)^c \tag{3}$$

where Δ*εep* is the range of elastic-plastic strain suffered by the component at the crack initiation area; *σ <sup>f</sup>* is the fatigue resistance coefficient; *ε <sup>f</sup>* is the fatigue ductility coefficient; *b* is the fatigue resistance exponent; *c* is the fatigue ductility exponent; *E* is the modulus of elasticity; and *Nf* is the fatigue life cycles.

The materials properties necessary to solve this probabilistic model are detailed in Table 2:

**Table 2.** Material properties for the probabilistic model.


#### **3. Results**

#### *3.1. Plates*

To obtain a correct clinical behavior, loads must be uniformly distributed throughout the four plates and transmit small stresses to the maxilla. Figure 9 shows the von Mises stress on the plates. Stress distribution along the plates is different because of the anatomical geometry of the maxilla, however, these differences on stress values are very small, as Figure 10 details.

**Figure 9.** The von Mises stress on plates in MPa.

In Figure 9, the maximum von Mises stresses appeared around the area where the prosthesis was adjusted to the plate and the body of the plate. The right HENGG-1 plate supported a stress of 185 MPa, while the same plate on the left had a maximum von Mises stress of 179 MPa. Stress on the right HENGG-2 plate was 168 MPa while in the plate placed on the left, it was 165 MPa.

According to Figure 10, the difference between the maximum von Mises stress in the HENGG-1 right and left plates was 3%, while the difference between the HENGG-2 right and left plates was 2%.

**Figure 10.** The von Mises stress values (in MPa).

#### *3.2. Maxilla*

All plates showed similar distribution patterns of maximum principal stress over the atrophic maxilla. The difference in the principal stress value between the four regions in contact with plates was 5%, with a mean maximum value of 80 MPa in the plates' region.

#### *3.3. Long-Term Behavior*

Failure probability of the situation detailed previously was obtained by employing the probabilistic methodology. The expected life computed was 1,819,235 ± 22.6 cycles. Then, the cumulative probability function was computed and represented in Figure 11.

**Figure 11.** Cumulative probability function.

Figure 11 relates the probability of failure associated with each cycle load. As shown in the previous figure, the probability was equal to zero until 1,300,000 cycles.

#### **4. Discussion**

The biomechanical behavior of CF@O plates on a completely edentulous and atrophic maxilla was evaluated in the present study by employing finite element methods.

For a few years, there has been a trend towards minimally invasive implant treatment in very atrophic edentulous jaws and maxillae. The purpose of these concepts is to make an implant

treatment with a shorter duration and smaller surgical risks [35]. The existence of insufficient bone can strongly influence the choice of the most appropriate rehabilitation in edentulous patients. There are several studies available in the literature that have employed different techniques to treat edentulous and atrophic maxilla, such as basal disk implants [36] or bone augmentation [37]. This study analyzed a new alternative based on titanium hybrid-plates. The accuracy of the results in numerical simulation studies depends on the precision of the model analyzed, the material properties, and the constraining conditions [38]. CT was used to model the geometry of the atrophic maxilla while the plates were provided by the manufacturer. A real clinical case with four titanium hybrid-plates in an atrophic maxilla was modeled and analyzed with the goal of knowing the biomechanical behavior of those plates.

This study had some assumptions and limitations. All materials were considered homogeneous, isotropic, and linearly elastic. Although these assumptions do not occur in clinical practice, they are common in finite element studies due to the challenges in establishing the properties of living tissues. These assumptions are consistent with other numerical studies [22,23,39]. In addition to these limitations, this work did not analyze the role of pterygoid implants as the goal was to study the biomechanical behavior of the plates. In that sense, ideal load distribution was considered.

The application of loads on the plates were supposed as an ideal transfer of loads from the prosthesis to that plate. If there is a good fit between the plate and the prosthesis, forces will be transmitted uniformly and as it was simulated and assumed in this study.

The ultimate strength in titanium grade II has been described as between 275 and 410 MPa and the ultimate tensile strength as 344 MPa. From these finite element analysis results, the maximum von Mises stress values in plates were lower than the ultimate strength [27]. The difference between the maximum von Mises stress between the distal plates was 3%, while the difference between the mesial plates was 2%. This difference was due to the different geometry of the maxilla in each area.

Küçükkurt et al. [40] compared the biomechanical behavior of different sinus floor elevations for dental implant placement. Under the condition of vertical loadings, von Mises stress in mesial implants were lower than our results in the plates in the case of lateral sinus lifting. However, the plates analyzed in this study obtained lower von Mises stress values than the prosthetic distal cantilever application and short implant placement. Regarding the distal implants, our plates obtained lower stresses than the prosthetic distal cantilever.

Ihde et al. [41] numerically analyzed baseplate implants with a vertical load of 114.6 N and a horizontal load 17.1 N and obtained a maximum von Mises stress of 400 MPa. Ihde et al. [42] detailed the von Mises stress in basal implants depending on the bone interface contact (BIC) degree. In this study, the maximum von Mises stress values were between 649 and 190 MPa. In both cases, the titanium hybrid plates used and analyzed in this study obtained lower stresses.

Kopp et al. [43] calculated the distribution of stress when basal implants in the mandible were loaded at two different stages of bone healing. They applied a load of 450 N located in the middle between the left molar and left canine implant and oriented in a vertical direction. Under these conditions, the von Mises stress in the basal implants was around 565 MPa.

The ultimate stress is an important value to understand the limits of the behavior of a material. According to physiological limits (ultimate stress), overloading in the cortical bone has been described as 170 MPa in compression and 100 MPa in tension [44]. Dos Santos et al. [45] detailed in their study that cortical bone resorption occurred when stress was higher than 167 MPa. Based on these limits, the values observed in this model were lower than those considered physiologic to bone tissue. As bone cannot be considered a ductile material, von Mises stress cannot be calculated in the maxilla. In this case, principal stresses have to be employed and calculated although some published studies have not used this kind of stress [39,45,46].

This is the reason why it is difficult to compare the results obtained in the atrophic maxilla of this study with the results provided in the bone in other published studies.

A good biomechanical behavior of plates is understood when a homogeneous stress is transferred to the bone. In this case, the maximum difference between the region of all four plates was 5%, meaning that the principal stress transferred from the plates to maxilla can be considered homogeneous.

Küçükkurt et al. [40] obtained a similar maximum principal stress in maxilla than our results in the case of short implant placement, and higher principal stress to our results in prosthetic distal cantilever application.

Clinical failures have generally been observed in the posterior maxilla region. Most of those failures were observed in bone types 3 and 4, with a highest probability of failures in bone type 4 [47]. According to the results obtained, the mean expected life in this case was 1,819,235 cycles. As Haug et al. detailed in their study [48], one year of in vivo service corresponds with, approximately, 200,000 cycles. Maló et al. obtained a satisfactory long-term outcome from patients with completely edentulous, severely atrophic maxillae supported by immediately loaded zygomatic implants alone, or in combination with conventional implants [49]. The same satisfactory results were detailed in Migliorança et al. who employed zygomatic implants placed lateral to the maxillary sinus and combined with conventional implants for a rehabilitation of the edentulous maxilla [50].

Further studies simulating these titanium hybrid-plates alternatives for atrophic maxilla and jaw that include dynamic forces that occur during chewing and consider the anisotropic and regenerative properties of bone are needed. Furthermore, some in vivo clinical trials are necessary to validate the model and to confirm the efficiency of this protocol. A numerical study of the combination of prosthesis-plates-implant under different functional conditions (bruxism and other parafunctions) just like the antagonist arcade is also necessary. Finally, a simulation of blood flow and bone regeneration around the plates is also necessary.

#### **5. Conclusions**

Based on the study results, it is possible to conclude that in terms of clinical application, the results indicate that titanium hybrid-plates proposed as an alternative to severe atrophic maxilla seems to have, from a mechanical point of view, a better behavior than conventional treatments such as prosthetic distal cantilever application and short implant placement. Titanium hybrid-plates distributed load to the maxilla with similar values as short implants but, with higher values than the prosthetic distal cantilever application. In any case, the resistance limits of bone and titanium were not exceeded. Long-term outcomes also seemed to be better than those clinical cases to treat atrophic maxilla. Finally, this technique can be considered as a viable alternative for atrophic maxilla rehabilitation, although more studies are necessary to corroborate the clinical results. As a clinical implication, this treatment allows the patient to be provided with functionally adequate prosthetic rehabilitation, which implies the recovery of their quality of life in a patient with a severe atrophy and, therefore, with an important challenge to a conventional implant treatment.

**Author Contributions:** M.P.-P. conceived, designed, and performed the analyses, evaluated the results, and wrote part of the paper. H.D. provided critical analysis. J.C.P.-F. provided critical analysis, interpretation of data, reviewed the literature, and wrote part of the paper.

**Funding:** This research was partially funded by Proclinic grant number A-285 and Instradent grant number A-274. Principal Researcher: Juan Carlos Prados-Frutos.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Corrosion Study of Implanted TiN Electrodes Using Excessive Electrical Stimulation in Minipigs**

#### **Suzan Meijs 1,\*, Kristian Rechendorff 2, Søren Sørensen <sup>2</sup> and Nico J.M. Rijkhoff <sup>1</sup>**


**\*** Correspondence: smeijs@hst.aau.dk

Received: 27 February 2019; Accepted: 25 March 2019; Published: 28 March 2019

**Abstract:** (1) Background: Titanium nitride (TiN) electrodes have been used for implantable stimulation and sensing electrodes for decades. Nevertheless, there still is a discrepancy between the in vitro and in vivo determined safe charge injection limits. This study investigated the consequences of pulsing implanted electrodes beyond the in vivo safe charge injection limits. (2) Methods: The electrodes were implanted for a month and then pulsed at 20 mA and 50 mA and 200 Hz and 400 Hz. Afterwards, the electrodes were investigated using electrochemical and analytical methods to evaluate whether electrode degradation had occurred. (3) Results: Electrochemical tests showed that electrodes that pulsed at 20 mA and 200 Hz (lowest electrical dose) had a significantly lower charge injection capacity and higher impedance than the other used and unused electrodes. (4) Conclusions: The electrodes pulsed at the lowest electrical dose, for which no tissue damage was found, appeared to have degraded. Electrodes pulsed at higher electrical doses for which tissue damage did occur, on the other hand, show no significant degradation in electrochemical tests compared to unused implanted and not implanted electrodes. It is thus clear that the tissue surrounding the electrode has an influence on the charge injection properties of the electrodes and vice versa.

**Keywords:** implanted electrodes; electrical stimulation; corrosion

#### **1. Introduction**

Titanium nitride (TiN) has been used for implantable electrodes for many decades, starting with cardiac pacing electrodes [1]. The demands on cardiac pacing electrodes increased when it was desired to sense the heart rhythm, in order to provide rate-adaptive pacing [2]. Despite the high voltages applied during cardiac pacing, the electrode polarization should remain low so that the heart signal can reliably be recorded [1,3]. Porous electrodes were highly desirable for that purpose [4] but the electrodes should also be biocompatible [1] and corrosion resistant [3].

At the end of the previous century, TiN also received interest as a material for neural stimulation and recording electrodes [5]. Neural stimulation and recording applications within this field include, among others, visual prosthesis [6], brain implants [7,8] and cochlear implants [9]. Initially, studies reported conflicting results [5,6,9], which was likely due to differences in the fabrication method [4]. The majority of studies, however, reported very favourable properties of porous TiN [5,6], which were due to its large surface area rather than specific material properties [4].

The performance of stimulation and recording electrodes can be evaluated using their safe charge injection limits (Qinj), charge storage capacity (CSC) and impedance. Qinj is evaluated by comparing the electrode polarization under pulsing conditions to the safe potential limits established using slow sweep cyclic voltammetry (CV). The safe potential window is typically defined by the potentials at

which water reduction and oxidation occurs. CSC is a measure of how much charge can be stored on the surface of the electrode and is measured using CV. The amount of charge available during fast pulsing, however, is typically much less than CSC. Impedance magnitude (typically at 1 kHz) can be used as a measure for battery consumption or recording performance. The lower the impedance, the better [10].

These properties are typically investigated under in vitro conditions in inorganic saline [5,6,9–20]. However, CSC, Qinj and the impedance spectrum differ under acute and chronic in vivo circumstances [4,7,8,21–28]. Qinj and electrode polarization have been reported to be significantly lower after implantation compared to in vitro measurements [7,8,23–27]. Moreover, they have been reported to decrease during the implanted period, when electrode failure does not occur [7,23–26].

TiN has long been known as a biocompatible [29–31] and corrosion resistant material [3,32,33], even under cathodic high voltage pulsing conditions [34]. Under anodic conditions, TiN oxidation reactions may occur, which primarily lead to passivation of these reactions until higher anodic voltages are reached [32]. At very high anodic voltages, TiN will eventually be degraded [34]. However, as Qinj is lower when implanted compared to in vitro [7,8,21–27], unsafe voltages may be reached during pulsing. The aim of this study was therefore to investigate whether implanted TiN electrodes would degrade during pulsing when Qinj measured in vivo was exceeded but Qinj measured in saline was not.

#### **2. Materials and Methods**

Four Göttingen minipigs were implanted with four working electrodes (electrode pins) and four large surface area pseudo-reference disk electrodes. Minipigs were selected because the subcutaneous adipose tissue is similar to adipose tissue in humans. The number of electrodes and pseudo-reference electrodes was chosen in order not to cause excessive discomfort to the animals and thereby also to increase the homogeneity in the results. The electrodes were made from Ti6Al4V and coated with porous TiN. The animals recovered from anaesthesia and were monitored for one month before the corrosion experiments were conducted. The work was carried out according to Danish and European legislation (ethical approval license no: 2014-15-0201-00268).

#### *2.1. Electrode Fabrication*

TiN coatings were deposited on electrode pins (6 mm2) made of a Ti6Al4V alloy and Ti disks (1000 mm2) by reactive magnetron sputtering on a CC800/9 SiNOx coating unit (CemeCon AG, Würselen, Germany). The coatings were sputtered from four Ti targets (88 × 500 mm2) with 99.5% purity in a Ar/N2 mixture atmosphere. The purity of the gases was 99.999% and the Ar/N2-flow was 300 sccm/350 sccm. The deposition time was 21,000 s. The electrodes underwent three-fold rotation during the coating process.

The electrodes coatings were investigated after deposition using samples taken from the same batch. Scanning electron microscopy (SEM) (Nova 600, FEI Company, Hillsboro, OR, USA) images were taken at magnifications ranging from 450× to 25,000× to get an overview of the surface and to investigate the surface structure of the electrodes in detail. A silicium sample coated in the same process was used to study the thickness, homogeneity and porous structure of the TiN coating using SEM. Images were recorded at a magnification of 40,000×.

An ethylene tetrafluoroethylene (ETFE) coated 35N LT wire (Heraeus, Yverdon, Switzerland) was crimped to the hollow end of the electrode pins. A polyether ether ketone (PEEK) body and silicone tines were produced using injection moulding to insulate the electrode pins. The tines were first glued to the PEEK body using a silicone adhesive. The PEEK body with tines was then glued to the electrode pins also using a silicone adhesive. Further details and figures of the electrode production can be found in Reference [24]. The electrodes were cleaned thoroughly before they were sterilized by an overdose of electron-beam processing.

#### *2.2. Surgical Procedure*

For electrode implantation, the animals were anesthetized using Propofol. Small incisions were made in the back and the electrodes were implanted using a custom-made implantation tool. This ensured that the electrodes were placed in tight pockets in subcutaneous adipose tissue, which promotes fast healing and ingrowth. Five electrodes were implanted into each pig, four of which were used for intense electrical stimulation and one electrode in each pig served as a control. Four counter electrode disks with a percutaneous wire were also implanted in each pig. The minipigs recovered from the procedure and were carefully monitored in order to detect and treat cases of infection. The electrodes were not used for one month until the pigs were anaesthetized again using sevoflurane to perform electrical stimulation. After the stimulation sessions were completed, the electrodes were carefully dissected from the tissue. The electrodes were extensively cleaned using demineralized water and alcohol, they were then rinsed and stored dry, so that they could be further investigated.

#### *2.3. Electrical Stimulation*

Electrical stimulation was performed using a DS5 (Digitimer, Hertfordshire, UK) per electrode. The device was shorted between the pulses using a custom-build set-up to prevent drifting of the baseline potential. Biphasic, charge balanced 200 μs square pulses were applied, cathodic first with an inter-phase interval of 40 μs during which no current was applied. Stimulation was performed for 6 h in total, divided into three 2-h sessions. Before, between and after these sessions, voltage transient measurements (VTM) were recorded for each electrode using a VersaSTAT 3 potentio-galvanostat (Princeton Applied Research, Oak Ridge, TN, USA).

Four stimulation paradigms were applied:


During pilot experiments, it was verified that the group 1 stimulation paradigm did not cause tissue damage after one week of implantation (see Figure S1). To cause electrode damage, we decided to increase 2 parameters: stimulation frequency and stimulation amplitude. The stimulation frequency was doubled, which was expected to cause electrode damage due to an increasing trend in the inter-pulse potential [35]. The stimulation amplitude was set to the maximum the DS5 can deliver, which was expected to cause electrode damage by increasing the electrode potential during stimulation. It was expected that group 2–4 protocols would result in tissue damage; therefore photographs were taken of the tissue surrounding the electrodes to document the amount of tissue damage. However, the focus of this study is corrosion and the electrodes were thus investigated more extensively using electrochemical and analytical methods.

During stimulation, the voltage transients were recorded every 30 min using an oscilloscope. From these voltage transients, the resistive drop after pulse cessation (IR-drop) was calculated as [10]:

$$\text{IR-drop} = \text{E}\_{\text{pulse\\_end}} - \text{E}\_{\text{pulse\\_end}+40} \tag{1}$$

where Epulse\_end is the recorded potential at the end of the cathodic pulse and Epulse\_end+40 is the potential 40 μs after pulse cessation (see Figure 1) [10]. Emc and Ema are the maximum cathodic and anodic voltage excursions after IR-drop is subtracted from the original voltage transient.

**Figure 1.** IR-drop, dV/dt, Emc and Ema are derived from the original data, while in the manuscript normalized data are presented. The data is normalized by subtracting the IR-drop and setting the pre-pulse potential to 0.

VTM before, between and after the stimulation blocks, was performed using the VersaSTAT 3 potentio-galvanostat using the same stimulation pulse.

The pulsing capacitance (Cpulse) was computed using the slope (dV/dt) of the voltage transient:

$$\mathbf{I}\_{\rm stim} = \mathbf{C}\_{\rm pulse} \cdot \frac{d\mathbf{V}}{dt} \tag{2}$$

where Istim is the stimulation current (1 mA while implanted and 5 mA in the electrochemical characterization). Qinj was calculated using the current (Imax) at which Emc or Ema reached the safe potential limits (−0.6 and 0.9 V vs. open circuit potential, respectively) [10,24]:

$$Q\_{inj} = \frac{\mathbf{I}\_{\text{max}} \cdot \mathbf{t}}{\mathbf{A}} \tag{3}$$

where t is the pulse duration (200 μs) and A is the geometrical surface area of the electrodes (6 mm2). When voltage excursions exceeded machine limits (±10 V), Imax was extrapolated from the highest current assuming a linear relation.

$$V\_{\rm ext} = V\_{\rm m} \left( 1 + \frac{\mathbf{I\_{ext}} - \mathbf{I\_{m}}}{\mathbf{I\_{m}}} \right) \tag{4}$$

where Vm and Im were the measured potential and current, respectively, and Vext and Iext were the extrapolated potential and current. When Vex reached the potential limits, Iext was used as Imax in (3). This method provided accurate results using data for which Imax was measured.

#### *2.4. Coating Characterization*

After explantation, all electrodes were characterized using SEM and energy-dispersive x-ray spectroscopy (EDX). The electrochemical properties were investigated using electrochemical impedance spectroscopy (EIS), CV and VTM. Two electrode groups were added to the 4 groups of active implants, therefore these measurements were performed on six electrode groups:


SEM (Nova 600, FEI Company, Hillsboro, OR, USA) images were recorded at magnifications varying from 450× to 10000× to obtain an overview of the surface and to investigate in detail the surface structure of the electrodes. SEM images of the not implanted control electrodes (group 6) were made, both to compare to the other electrode groups, as well as to investigate the uniformity of the coating after deposition. Further SEM analysis was carried out on a Si-wafer which was coated in the same process as the electrodes. The Si-wafer was mounted in a manner similar to the electrodes and the measured thickness is representative for the coating thickness on the electrodes. The advantage of using Si-wafer is that a cross-section analysis of the coating can be done easily. EDX (EDAX, AMETEK, Leicester, UK) spectra were made to investigate the chemical composition of the coatings after deposition and to determine whether the surface chemistry of the electrodes changed after having been implanted and after intense pulsing.

Electrochemical characterization measurements were performed in an electrochemical cell at room temperature using phosphate buffered saline as the electrolyte. The measurements were performed in a 3-electrode set-up, using the above mentioned porous TiN electrodes as working electrodes (0.06 cm2), a Ag|AgCl reference electrode (1.6 cm2) and a platinum foil counter electrode (50 cm2).

Solartron, Model 1294 in conjunction with 1260 Impedance/gain-phase Analyzer (Solartron Analytical, Farnborough, UK) were used to perform EIS measurements. Accompanying SMaRT software was used to run the measurements. A sinusoidal current was used at frequencies from 0.1 Hz to 100 kHz, with 10 measurements per decade. Three different currents (5, 10 and 50 μA) were used to ensure that the measurement currents were in the linear operation range of the electrode [36]. An integration time of 10 s was used to obtain a reliable and noise-free signal.

Cyclic voltammetry (CV) was performed by cycling the electrode potential was cycled between the safe potential limits (−0.6 and 0.9 V vs. Ag|AgCl) previously established for similar electrodes [24]. The sweep rates used for CV were 0.05, 0.1, 0.5 and 1.0 V/s. Ten cycles were made at each sweep rate, the last cycle was used for data analysis. The cathodic charge storage capacity (CSC) was derived from the CV by taking the integral of the CV below the zero-current axis [10].

VTM were conducted in the same manner as described above for the implanted electrodes, except the 3-electrode setup and the electrochemical cell were employed. The maximum charge injection limit (Qinj) and pulsing capacitance (Cpulse) were derived according to Equations (1)–(3).

#### *2.5. Statistics*

The data recorded during the 2-h pulsing sessions using an oscilloscope were filtered using a low-pass Butterworth filter (passband 5 kHz, stopband 15 kHz). Emc and IR-drop were then normalized to the first measurement (session 1, start). Emc was selected for statistical analysis to represent the electrode polarization and IRdrop, as a measure of the tissue resistance. Before, between and after the pulsing sessions, voltage transients were recorded using the VersaSTAT 3 (Princeton Applied Research, Oak Ridge, TN, USA). From these voltage transients Qinj and Cpulse were used to further quantify the electrochemical performance of the electrodes. A linear mixed model was used to statistically analyse the data. Parameters (group, session, time and combinations thereof) were added stepwise to the model, until adding another parameter did not make a significant difference to the model.

One-way ANOVA was used to investigate the electrochemical properties of the electrodes after explantation (electrochemical cell setup). The following electrochemical properties of the 6 different electrode groups were used for statistical analysis:


#### • Qinj

Significant findings are reported at p-values smaller than 0.05.

#### **3. Results**

All implantations were carried out without any complications. The animals recovered well from the surgery and no infections were observed during the month the electrodes were implanted.

#### *3.1. General Coating Characteristics*

In contrast to the well-known yellow-golden coloured TiN, the coatings on the electrodes had a brownish colour. To analyse the structure and chemical composition of the coating, the electrodes were studied in SEM and EDX. The overview SEM image (Figure 2a) shows a uniform coating on the electrode and in the corresponding EDX spectrum (Figure 2b) the expected peaks belonging to Ti and N are present. The quantification of the amounts of Ti and N from an EDX spectrum is difficult because the K-line of N and the L-line of Ti are very close. Here, numbers close to a 1:1 atomic ratio of Ti to N are found (note that weight-% is used in Figure 2b). The SEM images in Figure 3a clearly show a faceted structure, typical for TiN deposited at high pressure. The cross-section SEM image in Figure 3b shows the porous morphology of the coating as well. The coating thickness is approximately 6 μm.

**Figure 2.** (**a**) Scanning electron microscope (SEM) image of the coated electrode. The indicated region is the area over which EDX was performed. The scale-bar is 500 μm. (**b**) Typical energy dispersive X-ray spectroscopy (EDX) spectrum corresponding to the area indicated in (**a**).

**Figure 3.** (**a**) Top-view (magnification: 25,000×) and (**b**) cross-section SEM images (magnification: 40,000×) of TiN coating on Si-wafer. Scale bar: 1 μm.

#### *3.2. Changes in Electrochemical Properties during Intensive Pulsing*

The shorting part of the setup broke down during the last series of measurements. The last two stimulation sessions could therefore not be completed with one of the electrodes in the 20 mA—00 Hz group. The data obtained with this electrode after the breakdown was not used in the analyses.

The significant parameters of the statistical model for IR-drop were: Time, Session, Group × Session and Time × Session. For Emc, Group was an additional significant parameter of the statistical model. Figure 4 shows that the results for IR-drop and Emc were similar. Both IR-drop and Emc were significantly larger during session 1 compared to sessions 2 and 3 for electrode groups 2, 3 and 4. IR-drop and Emc were only significantly larger during session 1 compared to sessions 2 and 3 at the 30 and 60 min measurements. Figure 4a,b also show that IR-drop (for groups 2, 3 and 4) and Emc (all groups) were significantly larger after 30 and 60 min of pulsing compared to after 90 and 120 min of pulsing during session 1.

**Figure 4.** (**a**) IR-drop increased from baseline and then decreased during the first pulsing session for electrode groups 2, 3 and 4. (**b**) During session 2 and 3 (shown), an increase in IR-drop was seen from the start of stimulation, after which IR-drop remained stable. The IR-drop of group 2, 3 and 4 electrodes was significantly smaller during sessions 2 and 3 compared to session 1. (**c**) The same trend was observed for Emc of all electrode groups but to an even greater extent (notice the axis). (**d**) An increase in Emc was also observed from the start of stimulation during sessions 2 and 3 (shown). Emc of group 2, 3 and 4 electrodes was also significantly smaller during sessions 2 and 3 compared to session 1 but notice again the difference in the axis of IR-drop and Emc.

Emc of group 1 electrodes was significantly larger than Emc of group 2, 3 and 4 electrodes in all sessions. During session 2, Emc of group 2 electrodes was significantly smaller than Emc of group 1 and 3 electrodes and during session 3 Emc of group 2 electrodes was significantly smaller than Emc of group 1, 3 and 4 electrodes. Figure 4c,d show that IR-drop and Emc of all electrode groups was significantly smaller at the start of stimulation compared to after 30, 60, 90 and 120 min of pulsing during session 3. The same was found for session 2.

#### *3.3. Changes in Electrochemical Properties between Pulsing Sessions*

For both Cpulse and Qinj, the significant fixed effects were: Time, Group and Time\*Group. The results of the statistical analysis for Cpulse and Qinj were identical, except for a baseline difference between electrode groups observed for Qinj (p = 0.045). Qinj of group 2 electrodes (8.3 ± 2.4 <sup>μ</sup>C/cm2) was significantly smaller than Qinj of group 3 and 4 electrodes (15.8 ± 2.4 <sup>μ</sup>C/cm2).

Figure 5a,b show that Qinj and Cpulse, respectively, of group 1 electrodes did not change significantly. For group 2, 3 and 4 electrodes, Qinj increased significantly to values of 26.45 ± 2.7 <sup>μ</sup>C/cm2, 50.00 ± 2.4 <sup>μ</sup>C/cm<sup>2</sup> and 52.50 ± 2.4 <sup>μ</sup>C/cm2, respectively, after 6 h of intense pulsing. Cpulse of electrode groups 2, 3 and 4 increased significantly to capacitances of 54.1 ± 5.4 <sup>μ</sup>F/cm2, 97.1 ± 4.9 <sup>μ</sup>F/cm2 and 106.0 ± 4.9 <sup>μ</sup>F/cm2, respectively. Figure 5c,d show that the increase in Cpulse caused a decrease in electrode polarization. This decrease in electrode polarization led to an increased Qinj.

**Figure 5.** (**a**) After the first pulsing session Qinj was increased compared to before pulsing for electrode groups 2, 3 and 4. (**b**) The same was observed for Cpulse of electrode groups 2, 3 and 4. (**c**) Normalized voltage transients recorded before pulsing at 3 mA. (**d**) Normalized voltage transients recorded at 3 mA after 2 h of pulsing.

Electrode group 1 had a significantly smaller Qinj than all other electrode groups after the first pulsing session, which remained after the second and third pulsing session. Furthermore, electrode group 2 had a significantly smaller Qinj than electrode groups 3 and 4 after the first pulsing session. This difference also remained significant after pulsing sessions 2 and 3. The same group differences were observed for Cpulse.

#### *3.4. Electrochemical Characteristics after Explantation*

The results of the electrochemical characterization in phosphate buffered saline were largely consistent across measurements, as shown in Figure 6. Group 1 electrodes had a significantly smaller CSC at 0.05 and 1.0 V/s and a significantly larger impedance magnitude at 0.1 Hz compared to all other electrode groups. They also had a significantly smaller Qinj compared to group 2 electrodes and the control electrodes in groups 5 and 6. But the Qinj of group 1 electrodes was not significantly different from group 3 and 4 electrodes.

**Figure 6.** (**a**) The impedance magnitude was significantly larger for electrode group 1 compared to the other electrode groups, which was implanted and pulsed at the lowest electrical dose. (**b**) The cyclic voltammogram shows that the charge storage capacity of electrode group 1 was significantly smaller than the other electrode groups. (**c**) The normalized voltage transients at 5 mA show that the slope of the group 1 electrodes was larger than the slopes of the other electrode groups but no significant difference was found for Cpulse. (**d**) Qinj of group 1 electrodes was significantly smaller than Qinj of group 2, 5 and 6 electrodes.

#### *3.5. Coating Properties after Explantation*

SEM images (see Figure 7) showed that the electrode surfaces were intact after 6 h of intense stimulation. The coatings were all undamaged and had the same faceted structure, which is typical for TiN, as the electrodes had after deposition of the coating. Figure 7 indicates that there were no differences in chemical composition of group 1 and group 4 electrodes. The EDX spectra of all electrodes in all groups showed similar levels of titanium, nitrogen and oxide.

**Figure 7.** (**a**) Close-up SEM image (magnification: 10,000×) of an electrode in group 1 (20 mA—200 Hz), showing that the surface structure is still intact (scale bar: 10 μm). The same is true for (**b**), showing the detailed structure of an electrode in group 4 (50 mA—400 Hz, scale bar: 10 μm, magnification: 15,000×). (**c**) Overview SEM (magnification 450×) and corresponding EDX images of an electrode in group 1 show that the chemical composition of the electrode surface after pulsing was very similar to (**d**) the chemical composition of the electrode surface of an electrode in group 4. Also visualized by an overview SEM (magnification 450×) and corresponding EDX images of an electrode in group 4.

#### **4. Discussion**

Several previous studies [37–39] have shown that TiN coatings grown at high N-partial pressures become porous. These coatings have a brown colour that clearly distinguish them from the standard golden-coloured TiN used on tools and components. For use in electrode applications, it is essential to use the porous type of TiN coating as they have a high effective surface area, which in turn leads to a low impedance [3,4]. Cunha et al. [39] did a systematic study of the influence of N-content of the morphological structural and electrochemical properties of TiN coatings. In the case of high N-content (Ti:N ratio 1:1.34), these authors obtain results similar to ours regarding the coating morphology. The reason for the discrepancy in chemical composition (in our case we measure a Ti:N ratio 1:1 with EDX) could be that Cunha et al. [39] have determined the chemical composition using Rutherford Backscattering Spectrometry, which provides a more precise determination compared to EDX. Other studies have found that porous near-stoichiometric TiN coatings can be obtained just by adjusting the energy available during film growth [40].

Six groups of electrodes were used in this study, five of which were implanted and four of which were used for 6 h of intense pulsing at different electric doses:


During intense pulsing, IR-drop was derived from the voltage transients as a measure of tissue impedance and Emc was used as a measure of electrode polarization. Between the three 2-h pulsing sessions, Qinj was evaluated and from these voltage transients Cpulse was derived.

No significant changes were observed during intense pulsing for group 1 electrodes, receiving the lowest electrical dose. For group 2, 3 and 4 electrodes the IR-drop and Emc decreased and Qinj and Cpulse increased. The stability in pulsing properties of group 1 electrodes together with the fact that these electrodes received the lowest electrical dose, would intuitively lead to the expectation that these electrodes did not corrode [4,7,8,23–26]. For group 2, 3 and 4 electrodes, on the other hand, it could be expected that corrosion may have occurred, even though the observed electrochemical changes are favourable in the light of pulsing capability (higher charge injection, lower electrode polarization). Passivation at high anodic potentials was the main expected corrosive reaction [33,34], leading to decreased pulsing capability. Excessive bubbling due to water reduction [34], however, may lead to cracking of the coating. This could increase the surface area and thereby lead to an apparent increase Qinj. The results of the electrochemical characterizations after explantation showed that group 1 electrodes had significantly deteriorated electrochemical properties compared to all other electrode groups. Group 2, 3 and 4 electrodes, on the other hand, had no different electrochemical properties after explantation than the two control groups (5 and 6).

Analytical investigations could neither confirm nor reject the electrochemical results. SEM images show that all coatings seemed to be intact. EDX spectra did not reveal differences between the harshest and mildest stimulated electrodes either. However, it must be noted that it is difficult to distinguish between the oxygen and nitrogen signal using EDX because the K-peaks of the two elements are quite close. As oxidation of the coating could have occurred, other analytical methods have been attempted (X-ray photoelectron spectroscopy and Time-of-Flight Secondary Ion Mass Spectrometry) in order to detect any differences in oxygen amounts. Preliminary results were unsuccessful, mainly because the geometry and size of the electrode is quite challenging in both the experimental set-ups. It is, however, very unlikely that the coating oxidized without showing any signs of damage. Norlin et al. [34] show SEM images porous TiN electrodes after anodic pulsing, which are severely damaged. For the current electrodes no signs of damage were found using SEM and EDX analysis.

Corrosion studies of stimulation electrodes have been performed extensively in saline [3,34,41–43]. Some studies found that the damage threshold is exactly at the water window limit [43], others suggest that the water window limit may be too conservative under pulsing conditions [41,44] and yet others suggest that corrosion may occur even within the limits of water reduction and oxidation [42]. TiN has been most extensively investigated by Norlin et al. [3,34,45]. Their pulsing study was carried out using 700V pulses of both anodic and cathodic polarity [34]. As expected, TiN showed severe corrosion upon anodic pulsing but was stable when cathodic pulses were applied. The voltages recorded in this study during constant current pulsing were, however, a factor 15–30 smaller. In a later study, the electrodes were aged using more conservative voltages (−3 and 1 V vs. Ag|AgCl), corresponding to 4 months of use based on the charge passed [3]. TiN proved very stable, which was expected, as high anodic voltages were avoided. In the current study, very high anodic voltages were also avoided by using a cathodic first stimulation paradigm. Anodic voltages between 1 and 3 V (vs. open circuit potential) were observed with the highest voltages in the 50 mA groups (groups 3 and 4). No signs of corrosion were observed for the electrodes in those groups, while corrosion of the anodically pulsed samples by Norlin et al. [34] was obvious in SEM images.

It has also been shown before that safe limits obtained in inorganic solutions do not necessarily apply to electrodes in protein containing solution [46] and implanted electrodes [47,48]. It was therefore concluded that proteins must protect the electrode surface against corrosion [46,48,49]. However, Qinj was never measured in vivo for these electrodes and it is therefore unclear whether or not it was exceeded [47,48]. The in vitro water window limits for platinum were not exceeded in either of the studies but in both studies corrosion was observed nevertheless [47,48]. Shepherd et al. [48] argue, however, that corrosion was not stimulation-induced but due to production failures. In the current study, Qinj (measured in vivo) was exceeded for all stimulated electrode groups (1, 2, 3 and 4) during all stimulation sessions. However, group 1 and 2 electrodes were pulsed at approximately 20% of Qinj as measured in saline and groups 3 and 4 electrodes were pulsed at approximately 50% of the in vitro Qinj. Robblee et al. [47] stimulated their electrodes at approximately 5% and 30% of Qinj in vitro. They observed platinum dissolution for all electrodes pulsed at 30% of Qinj (in vitro) and less for electrodes pulsed at 5%. Shepherd et al. [48] stimulated the electrodes at 5-10% of Qinj in vitro but concluded that the observed corrosion was not stimulation-induced. This makes it obvious that in vitro safe limits cannot be applied in vivo. However, it does not rule out that limits measured in vivo using techniques developed in vitro may be too conservative.

Interestingly, we found that corrosion most likely occurred for the electrodes pulsed at the lowest electrical dose (group 1 electrodes; 20% of Qinj in vitro and 200 Hz). As group 2, 3 and 4 electrodes showed no signs of corrosion in the electrochemical characterizations after explantation, the occurrence of corrosion seems not only dependent on the electrode potentials or charge delivered. We suspect that the occurrence of corrosion is not only medium dependent (organic vs. inorganic, basic vs. acidic) but also tissue dependent. Although tissue damage was not the focus of this study, it seems to play an important role. No tissue damage seems to have occurred for electrode group 1, while tissue damage with increasing severity occurred for electrode groups 2–4 (see Figure S2). The lack of tissue damage for electrode group 1 is confirmed by the lack of change in IR-drop, which is representative of tissue impedance [10]. The same amount of charge density per phase was injected for electrode groups 1 and 2 but due to the increased frequency tissue damage is likely to have occurred in group 2 [50,51]. Tissue damage obviously occurred in electrode groups 3 and 4 (see Figure S2). Based on the IR-drop data, it appears that the tissue was damaged during the first hour of the first pulsing session for electrode groups 2, 3 and 4. It seems that a new electrode-tissue interface was formed that remained during pulsing session 2 and 3. This new electrode-tissue interface allowed for more charge injection, as Qinj and Cpulse were significantly increased after the first pulsing session compared to before pulsing. And although the electrodes were still pulsed beyond the increased Qinj in vivo, the formation of a new electrode-tissue interface and corresponding increase in Qinj may thus have prevented corrosion.

The electrode-tissue interface appears to play a very important role with regards to the occurrence of corrosion. These results can therefore only be applied to stimulation electrodes implanted in adipose tissue, like ours [23,24,52] and like Bion [53] for example. They cannot be applied to implants in the brain [47], the cochlea [48] or the blood stream [1,2,4]. Furthermore, our electrode is a macro-electrode (0.06 cm2). There are indications that different charge injection limits apply to smaller microelectrodes [51]. These results might therefore not apply to microelectrodes. Lastly, as it is challenging to work with larger animals, such as minipigs, the number of animals is low compared to rodent studies for example. The results, however, are consistent across measurements and rather homogeneous within the electrode groups and were thus statistically significant.

With the recent increase in investment in "electroceuticals," the development of novel, smaller and more sophisticated implants may be anticipated [54]. It is therefore more important than ever before to establish safe limits that apply to these specific implants [51]. We show that this is not only relevant in the light of tissue damage but also with respect to corrosion and long-term electrochemical performance of the implants. In the light of tissue damage due to corrosion, TiN appears to be a very suitable material for implants. There seems to occur no dissolution of the material [55], like with Pt [46] and IrOx [43]. As long as very high anodic potentials are avoided [34], we show that no corrosion occurs even after almost 9 million pulses. When corrosion does occur, its product (a passivation layer) remains attached to the electrodes and is not harmful to the tissue [33,55].

#### **5. Conclusions**

It was long suspected that in vitro safe limits established for implantable electrodes may not be applicable in vivo, which we confirm here. We also show that the type of tissue in which the electrode is implanted has an influence on safety limits. Biocompatibility and corrosion resistance cannot be viewed as two separate properties of implantable stimulation (and sensing) electrodes. Tissue responses influence the electrochemical behaviour of implanted electrodes and use of the electrodes influences the tissue surrounding the electrodes. It is therefore of great importance that safe limits are established for each electrode depending on the tissue in which it will be implanted.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2075-4701/9/4/389/s1, Figure S1: (a) Tissue around the tip of an electrode stimulated at 20 mA-200 Hz after 1 week of implantation. The tissue was stained using haematoxylin and eosin (H&E), which is the most commonly applied stain in medical diagnostics. Some inflammatory cells can still be observed but capsule formation has begun to take place. (b) Cells around the silicone part of the electrode appear very similar to those around the electrode tip, indicating no signs of stimulation-induced tissue damage. Figure S2: (a) The tissue around electrodes in group 1 showed no signs of tissue damage upon sacrifice. (b) The tissue around electrodes in group 2 showed some redness around the electrode tip, which likely is due to tissue damage. (c) The tissue around the electrode tips of electrodes in group 3 showed obvious tissue damage but the tissue around the insulated parts was unaffected. (d) The tissue around the electrode tips of electrodes in group 4 showed even more extensive tissue damage and bleeding. Nevertheless, the tissue around the insulated parts was unaffected.

**Author Contributions:** Conceptualization, S.M. and N.R.; methodology, S.M., K.R., S.S. and N.R.; writing—original draft preparation, S.M.; writing—review and editing, S.M., K.R., S.S and N.R.; project administration, N.R.; funding acquisition, N.R.

**Funding:** This research was funded by the Danish National Advanced Technology Foundation.

**Acknowledgments:** The authors thank Neurodan A/S, a member of the Ottobock group, for supplying the electrodes used in this study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Metals* Editorial Office E-mail: metals@mdpi.com www.mdpi.com/journal/metals

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18