Floating Light-Activated Micro Electrical Stimulators Tested in the Rat Spinal Cord

Abstract

Microelectrodes of neural stimulation utilize fine wires for electrical connections to driving electronics. Breakage of these wires and the neural tissue response due to their tethering forces are major problems encountered with long term implantation of microelectrodes. The lifetime of an implant for neural stimulation can be substantially improved if the wire interconnects are eliminated. Thus, we proposed a floating light-activated micro electrical stimulator (FLAMES) for wireless neural stimulation. In this paradigm, a laser beam at near infrared (NIR) wavelengths will be used as a means of energy transfer to the device. In this study, microstimulators of various sizes were fabricated, with two cascaded GaAs p-i-n photodiodes, and tested in the rat spinal cord. A train of NIR pulses (0.2 ms, 50 Hz) was sent through the tissue to wirelessly activate the devices and generate the stimulus current. The forces elicited by intraspinal stimulation were measured from the ipsilateral forelimb with a force transducer. The largest forces were around 1.08N, a significant level of force for the rat forelimb motor function. These in vivo tests suggest that the FLAMES can be used for intraspinal microstimulation even for the deepest implant locations in the rat spinal cord. The power required to generate a threshold arm movement was investigated as the laser source was moved away from the microstimulator. The results indicate that the photon density does not decrease substantially for horizontal displacements of the source that are in the same order as the beam radius. This gives confidence that the stimulation threshold may not be very sensitive to small displacement of the spinal cord relative to the spine-mounted optical power source.

I. INTRODUCTION

Electro-magnetic, optical, or acoustic energy can be wirelessly transferred to neural tissue to stimulate the neurons remotely [1, 2] or modulate their activity [3]. These direct methods of wireless neural stimulation are not commonplace in neural prosthetic applications yet due to the high levels of energy needed and the bulky extracorporeal equipment required. One way to reduce energy requirements is to use an implantable microstimulator that can harvest the energy first and then deliver it to the local tissue whenever the stimulation is needed. Energy storage and pulse shaping can also be achieved through electronics that can be incorporated into this wireless device. However, any additional component that can increase the device size should be avoided for microstimulation applications in the CNS because of concerns with tissue trauma. A completely passive device that harvests the energy and instantaneously converts to electric pulses for stimulation would also reduce the risk of failure because of its simplicity of operation.

Our group has been investigating a floating light activated micro-electrical stimulator (FLAMES) that contains passive photodiodes for wireless activation of neural tissue using near infrared (NIR) light for energy transmission [4, 5]. In the envisioned paradigm, the micro stimulator is implanted into the neural tissue at the targeted site in the CNS, e.g. the spinal cord gray matter, typically a few millimeters below the pial surface. The laser pulses are sent through a multi-mode optical fiber placed near the micro stimulator to activate it. The tip of the optical fiber is inserted through a hole into the vertebrae just above the micro stimulator, but located outside the dura matter (Fig. 1). Therefore, the distance that the laser beam has to travel is on the order of a few millimeters through the neural tissue before reaching the microstimulator. The laser source and control electronics are implanted at a distant site, e.g. the subclavicular area, that is most convenient for transcutaneous charging of batteries and programming of the pulse parameters using RF telemetry.

Fig. 1.

The envisioned implantation paradigm for chronic microstimulation of the spinal cord gray matter using a FLAMES and an optical fiber that delivers the stimulation energy in the form of NIR light through the neural tissue.

Wireless stimulators have been reported by other groups [6–8]; an optically activated, yet tethered, micro stimulator was shown to stimulate the rat sciatic nerve [6]. This design apparently does not intend to address the problems of interconnects since it is attached to a fiber optic tail that replaces the wires. Optical energy is delivered via the optical fiber rather than being transmitted through the neural tissue, as it is done with the current design. An ultrasonically powered wireless device was also proposed for nerve stimulation [7]. The stimulator length is 8mm in a recently reported version and thus may be too large for CNS applications. BION^™ stimulator is an example of a wireless stimulator that is controlled by radio-frequency electromagnetic waves and designed to be used in the peripheral nervous system [8]. Because the telemetry and power circuitry are incorporated into the stimulator, the device size is on the order of centimeters.

The main objective of this study was to investigate the feasibility of the FLAMES approach for intra-spinal microstimulation (ISMS), which has been investigated as a potential technique for recovering function in a number of conditions caused by spinal cord injury (SCI). The spinal cord includes specialized neural circuits that can autonomously generate coordinated muscle activity [9]. It was shown that a few electrodes implanted in the cat spinal cord within the lumbosacral region can produce locomotion [10–12]. Losing somatic and visceral functions, such as limb movements and bladder voiding, are some of the disabilities that result from SCI. Feasibility of ISMS for limb and bladder control has been demonstrated in animal models [13]. Breakage of interconnects [14] and chronic tissue response induced by the micromotion of electrodes [15] is a major impediment before ISMS can be pilot studied in human subjects [16]. Electrodes made with flexible wires have been demonstrated to last for several years in the cortex [17, 18]. However, the interconnect longevity in the feline spinal cord is rarely more than a year [19]. In an attempt to test the feasibility of ISMS for bladder control in cats, ~20% of the electrodes implanted in the sacral spinal cord became non-functional due to lead failures after 3 months of implantation [19]. In another study where the cat lumbar region was microstimulated to generate leg movements, only 67% of the electrodes remained functional after 6 months of implantation [20].

The human spinal cord experiences substantial translational and rotational displacements that make chronic implantation of micro wire electrodes very challenging. For instance, the log-roll rehabilitation technique developed for traumatic spinal cord injury subjects generated 7.3±5.8° of flexion/extension and 7.9±9.1° of axial rotation of the cervical cord as tested in cadavers [21]. A floating, wireless microstimulator can be an important step towards the development of such a system for use in human subjects.

Single-channel wireless stimulators may resolve or alleviate these problems by eliminating the tethering forces due to interconnects, which is the primary source of chronic tissue reaction [22]. With no wires attached, the implanted device would float and move together with the surrounding neural tissue as it goes through deformations, and translational and rotational displacements thereby substantially reducing the shear forces at the device-tissue interface.

Various sizes of FLAMES were fabricated in the sub millimeter range with integrated gallium arsenide (GaAs) p-i-n photodiodes. The current-voltage (I–V) curves were obtained to show the photodiode characteristics of the devices. The gold contacts were coated with Poly 3,4-ethylenedioxythiophene (PEDOT) to improve the charge injection capacity. The devices were placed in a saline solution to measure the stimulation voltage near the cathode in a conductive medium. The FLAMES were then tested in the rat cervical spinal cord for activation of forelimb musculature. The forces generated through ISMS were measured from the ipsilateral forelimb with a force transducer.

II. METHODS

A. Device Fabrication

The III–V semiconductor heterojunction structure in Fig. 2A was designed and then grown at the Institute of Electronic Materials Technology, Poland on a GaAs substrate. The tandem device was designed to have two p-i-n photodiodes connected through a highly doped tunneling junction. The top part of the device, the optical blocking layer, contains a one-micron thick film of Al₁₀Ga₉₀As to absorb the unwanted lower wavelengths and prevent them from reaching the photodiodes.

Fig. 2.

Fig. 2A. Al_xGa_(1-x)As heterojunction structure used in the fabrication of the FLAMES devices. B. (a) side view, (b) front view, (c) top view of FLAMES device after etching and thin film depositions. C. An annotated micrograph of a type C device showing the layout at the tip of a 3.5mm shank (an extension of the substrate for handling the device during testing). The scale bar is 100μm.

Aluminum gallium arsenide/gallium arsenide (AlGaAs/GaAs) FLAMES were fabricated in different sizes. Die pieces of 1×1cm were diced from the custom AlGaAs/GaAs 50.4mm wafer. S1813 positive photoresist (Shipley, Massachusetts) was used to pattern the devices for all steps. Mesa etching was utilized to micromachine the custom wafer in Fig. 2A into the vertical structure shown in Fig. 2B. Selective wet etching was used to reach the doped n layer by utilizing AlAs etch stop. Wet etchant for etching to p-type layer was a 25:5:1 deionized (DI) H₂O/phosphoric acid/H₂O₂ solution. The p metal layer was deposited with Pt: Ti: Pt: Au: 100: 400: 100: 1500Å and annealed at 400°C for 1 minute. For etching to n-type layer, a selective etchant was prepared by mixing 5.5ml H₂O₂ and 50ml citric acid solution (1g citric acid dissolved in 2g DI H₂O) to etch GaAs and Al_xGa_1-xAs layers and stop at the AlAs etch stop. It is very important to remove the thin surface oxide layer of the cap layer before etching. This was achieved by dipping the device in 15:1 DI H₂O/buffered oxide etch solution for 10 seconds. Then, a short dip in 15:1 H₂O/buffered oxide etch (BOE) removed the AlAs layer. The n metal deposition consisted of Ge: Au: Ge: Au: Ni: Au (60: 100: 100: 240: 100: 1500 Å) layers, which was then annealed at 430°C for 1 minute. Finally a 300nm silicon nitride layer was deposited with PECVD and patterned. The top silicon nitride layer was then etched with CF₄ plasma. The devices were released using a rotary dicing saw and a backside etch to thin the devices until they were approximately 100-to-150μm thick. Finally, the devices were cleaned using a dilute hydrofluoric acid bath and a long plasma oxygen descum. A micrograph of a fabricated device is shown in Fig. 2C.

B. Probe Station Measurements

Fabricated devices were optically characterized using a focused output from a multimode fiber that was imaged on the active area of the p-i-n photodiode. The device was placed on a mechanical stage that was adjusted to ensure that the optical beam was contained within the active area. BeCu coated tungsten probe tips (T-4 125-BeCu, GGB Industries, FL) were affixed to three-axis micromanipulators and connected to a semiconductor parameter analyzer (HP 4156A) where low frequency I-V measurements were performed. The devices were tested under two different optical powers, 0μW and 30μW with an 856nm wavelength laser diode.

Two photodiodes were vertically connected in series to increase the output voltage. The long metal connections on the shank devices were not insulated from the semiconductor material creating a resistive path for carriers. This resulted in a lower turn-on voltage than expected for two photocells in series. Non-shank devices did not exhibit this reduction in voltage, but the shanks were necessary for securing the devices to a micromanipulator during in vivo testing. The optical blocking layer on top reduces the efficiency of our devices. This layer was originally included in the design to test the idea of wavelength selectivity, but not further investigated in this report. Thus, our series photodiodes exhibit a relatively modest conversion efficiency of 11% (in agreement with simulations), which should typically be above 40% without the blocking layer [6].

C. Volume Conductor Measurements

PEDOT was deposited on the gold stimulation contacts using electrochemical polymerization with 10mM of PSS in an EDOT solution at 1.5 mA/cm² and allowing sufficient time for the contacts to turn black [23]. Overcoating was avoided to prevent flaking off the PEDOT film. The stimulator was secured at the bottom of a petri dish filled with normal saline (0.9%) solution diluted ten times to simulate neural tissue impedance (~600 ohms-cm). A tungsten micro-electrode was placed immediately above the cathode (~10–30 μm), without making direct contact, to record the voltage field generated by the FLAMES. A high input-impedance buffer amplifier (TLC2274, Texas Instrument) was utilized in this part and a large Ag/AgCl electrode as a reference. The stimulation voltage was measured in response to a single laser pulse at 830 nm (PW = 0.2 ms). The laser source (DLS-500-830FS-100, StockerYale, Canada, 74 mW) and acquisition of signals into a computer was controlled through MATLAB. The laser power was varied from zero to maximum to study the voltage field near the contacts in the medium.

D. In Vivo Testing

In six Sprague-Dawley rats (350–500g), anesthesia was induced with intraperitoneal injection of ketamine (80mg/kg) and xylazine (12mg/kg) mixture diluted with saline. Further doses of ketamine were administered to maintain the anesthesia as needed. Marcaine (0.2mL) was injected at the site of incision as a local anesthetic. The rectal temperature was continuously monitored and maintained between 36–37°C using a temperature regulated heating pad. The head was stabilized in a stereotaxic frame and the spinal cord was immobilized from the spinous process at T1. Laminectomy was performed at C5-C7 and the dura was removed. The spinal cord was kept moist using warm saline. All procedures were approved and performed in accordance to the guidelines of the Animal Care and Use committee, Rutgers University, Newark, NJ.

Three different device sizes with a top layout as shown in Fig. 2C were tested in this study. The largest device, type A, had cathodic and anodic areas of 11,500 μm² each, active area of 28,500 μm², and a total device size of 200×700 μm. These values for Type B and C devices were 8,500 and 5,500 μm², 25,500 and 17,500 μm², and 170×600 and140×500 μm respectively. Figure 3 shows the preparation used for intraspinal stimulation in anesthetized animals. The FLAME stimulator was inserted into the cervical spinal cord at an angle of 30 degrees (from the normal) near the left or right dorsal root entry zone and reaching the ventral horn on the ipsilateral side using a 3-axis micromanipulator (inset). The laser was placed 13.5cm above the cord using another micromanipulator and the built-in lenses were adjusted to form a beam with a circular footprint of 0.56 mm in diameter at the cord surface aiming at the photodiode from above. A train of NIR pulses (0.2 ms pulse width, 50Hz, 500ms duration) was sent to the device through the tissue. The power of the laser source was gradually increased until twitches in the forelimb were observed. The power at that level was considered the threshold (P_th). Then, the laser power was increased incrementally while the forces generated via spinal cord stimulation were measured with a transducer attached to the ipsilateral hand. The photodiode currents at threshold power were also measured at various depths from all devices (A, B, and C) using a current amplifier, which short circuited the device output to zero volts (hence no stimulation), preventing the current flow through the internal diode.

Fig. 3.

The intraspinal cord microstimulation setup for in vivo testing of FLAMES in rats. The FLAME stimulator is located at the tip of a shank that is inserted into the cervical spinal segment C6. The instantaneous power and the pulse width of the laser pulses are controlled via the computer. The forelimb extension forces generated by the ISMS are recorded by a force transducer attached to the ipsilateral hand.

The laser beam used in this study had a Gaussian profile with a circular footprint. The radius was taken a $\sqrt{2}\sigma$ of the Gaussian profile, which is the distance where the intensity drops down to 1/e² of its peak in 2D plane. This diameter corresponds to the diameter of a flat profile circular beam with the same peak power density and the same total power as the Gaussian profile laser. The instantaneous power was controllable and varied linearly with the control signal, as specified by the manufacturer. The incident power density at the cord surface (in mW/cm²) was calculated accordingly for a train of pulses using the equation below:

$$\mathit{power}\_\mathit{density}=\frac{P}{A}\times PW\times f$$ (1)

where P is the peak power of the laser pulse controlled by the computer (mW); A is the cross-sectional area of an equivalent, uniform profile beam at the cord surface (cm²); PW is the pulse width (sec); and f is the frequency(Hz).

III. RESULTS

A. Bench Top Measurements

Devices were tested using 856nm excitation from a laser diode. Figure 4 shows the measured I–V characteristics for the three devices with and without illumination. All three devices have similar open circuit voltages (V_OC) and short circuit currents (I_SC). When used as a stimulator, the device operates in the photovoltaic mode. That is, depending on the load resistance, the device will operate between the short circuit current and open-circuit voltage. There was no evidence of breakdown for a reverse bias of 30V. Responsivity was about 0.1A/W for all devices.

Fig. 4.

I–V measurements obtained using a bench top optical probe station.

B. Volume Conductor Measurements

The voltage field recorded immediately above the cathode in saline is shown in Fig. 5A for Device C. The peak voltage ~30μm above the cathodic contact was around 110mV and stayed flat for the duration of 0.2ms pulse. This suggests that the contact-medium interface has a sufficiently large charge injection capacity to maintain a constant current. The test was conducted at various frequencies within a range of 10–100Hz. The peak voltage was stable and the interface was discharged completely during the off cycles through the leakage currents of the photodiodes [24]. Figure 5B shows the voltage field in saline immediately above the cathode of Devices A, B, and C as a function of laser power. The voltage exponentially increases and then plateaus at a laser power where the internal pn junctions start limiting the device voltage. The voltage pulse amplitude was very sensitive to the electrode tip distance from the contact. Vibrations of the Tungsten recording electrode induced some variations into the plots.

Fig. 5.

Volume conductor measurements. A: The waveform recorded ~30μm above the center of the cathode (of Device C) as a response to a single laser pulse. B: The field potential recorded immediately (~10 μm) above the cathodes with respect to a Ag/AgCl electrode of Devices A, B, and C as a function of incident light power. Note that this is the instantaneous incident power density (not multiplied by the duty cycle, which is PW x f=1%) at the surface of the volume conductor.

C. In Vivo Testing

Elbow extension was the type of movement typically generated in C6 by stimulating the most ventral locations in the cord. The device tip and thus the cathodic contact was inserted about 2.35 mm below the dorsal pial surface to generate this forelimb movement. Figure 6A shows the reproducibility of the vertical component of the force generated during a 0.5s on-1.5s off cycle for 15s in one of the animals. The force in each cycle is fused and stable at 50Hz train frequency without any signs of fatigue (Fig. 6B). As control, the muscle contractions were completely blocked everytime an opaque object was placed in the path of NIR light.

Fig. 6.

Vertical component of the elbow extension force recorded with a transducer attached to the ipsilateral hand in rat 3: (A) for a series of 0.5s ON-1.5s OFF cycles. Each tetanic force is generated by a train of 50Hz, 0.2ms pulses. (B): A sample cycle expanded from the plot in A (marked with *) to demonstrate tetanus.

The forelimb extension forces are plotted in Fig. 7 collectively from all experiments as a function of incident laser power density at the cord surface. The largest forces produced with Devices A, B, and C were around 0.82N, 0.8N, and 1.08N respectively. The smallest and largest stimulation current values injected into the tissue were around 6±2.5 and 120μA for all devices respectively. The variations in the plots are due to slightly different implant locations in each animal.

Fig. 7.

Forelimb extension forces generated by intraspinal microstimulation at 50Hz (PW=0.2ms) as a function of incident laser power density calculated at the spinal cord surface according to Eqn. (1).

Force measurements are clustered into two groups. One group is in a low-threshold area and the other is in a high-threshold area. The two areas show two distinct behaviors. The low-threshold area has a larger slope where the force reaches the threshold at relatively small laser powers but elicits larger forces. The high-threshold area has a smaller slope where in some cases nine times higher power was required to reach half the force generated in the low-threshold area (e.g. Rat 4-Device B vs. Rat 5-Device A). In rat 4, data from both areas was obtained for comparison. The electrodes were moved a few hundred micrometers up to repeat the experiment into the high threshold area. In Rat 5, the low-threshold area behavior was also seen but the device tip had to be moved a few hundred micrometers above to avoid fatigue.

An example strength duration curve, which we define as the threshold laser power as a function of pulse width, is shown in Fig. 8 from one of the trials (dash line). The microstimulator was located in the low-threshold region in the cord as identified in Fig. 7. Threshold power decreases exponentially by the pulse width as expected in a strength-duration curve. Only a few milliwats suffice for activation with pulses longer than 0.2ms. The solid line in the figure is the curve fit generated using Lapicque’s equation for strength-duration curve on intracellular stimulation of neurons [25]. The chronaxie time and rheobase power are found as 0.27ms and 1.5mW respectively from the experimental data, which agrees well with the theoretical curve fit.

Fig. 8.

The strength-duration curve results obtained with Device C in rat 4 (dash line). The solid line is the curve fit using Lapicque’s strength-duration equation. The pulse width was varied from 30μs to 1ms while searching for the threshold laser power needed to generate smallest arm twitches.

The threshold power was investigated as the laser source was moved away along the spinal cord (Fig. 9). The vertical axis is in multiples of the threshold power that was measured when the beam was aligned with the active area of the microstimulator (P_th). The threshold power had to be increased 1.15±0.12 times when the source moved 250μm away from the center. Similarly, a 4±0.37 times higher power was required to produce the threshold arm movement when the source was displaced by 1mm.

Fig. 9.

The threshold laser power is shown in multiples of P_th as a function of the laser source displacement from the center of the device active area. The devices were implanted at ~2.35 mm below the pial surface. Data represents the average of eight trials from 4 rats using all three types of devices in most animals.

IV. DISCUSSION

A. NIR Exposure Limitation

NIR exposure safety limit is given only for the skin and retina by American National Standards for Safe Use of Lasers (ANSI Z136.1-2007). Radiating the neural tissue with light can cause temperature elevation that can be harmful to the tissue if the temperature increases by more than 1°C [26]. Computer simulations suggested that in order to keep temperature elevation under 0.5°C in the neural tissue, NIR exposure needs to be limited to 325 and 250mW/cm² for gray and white matter respectively [27]. The maximum incident power density used in this study was 75mW/cm², which is only a fraction of our predicted maximum allowable exposure.

Optical properties of tissue determine the penetration depth of light. The scattering coefficient is the most important factor that determines the number of photons available at any given depth. It is important to note that the scattering coefficients values reported in the literature for human and animal neural tissues vary within a large range [28–30]. The human spinal cord is also about two and a half times thicker (sagittally) than the rat spinal cord at the cervical level (5.8mm at C6) [31]. ISMS of human spinal cord may be more demanding because the targeted sites for stimulation may be deeper in the cord. Considering all these species differences, simulations suggested that microstimulation of human spinal cord is feasible using comparable sizes of FLAMES to the ones in this study [27].

The microstimulators were efficient enough to convert the relatively small amount of NIR light available at a depth of 2.35 mm inside the neural tissue to a sufficient electrical current for ISMS. This suggests multiple FLAMES can be activated next to each other without exceeding the safe level of NIR exposure. On the other hand, it should be mentioned that in chronic applications the amount of current required for stimulation is higher due to encapsulation of electrodes with connective tissue. This warrants further testing of FLAMES with long-term implants.

In the envisioned chronic stimulation paradigm (Fig. 1), the laser source will be fixed to the vertebral bone and therefore the devices may move in and out of focus as the spinal cord is moving inside the vertebrae. The experimental setup did not allow us to test for changes in the threshold power for rotational displacements. The longitudinal translations of the source suggested that the threshold increases by four times for a displacement of about four times the laser beam radius. Increasing the beam size can also reduce the sensitivity to source displacement. As a trade off, however, a larger tissue volume would be radiated by the laser beam.

B. Microstimulator Sizes

Three different sized devices were tested in this study. One of the goals of this research is to minimize the device size while generating sufficient current for stimulation. Minimizing the stimulator size is crucial in order to reduce the neural tissue replaced by the implant and the long-term immunological response. Depending on the desired volume of tissue to be stimulated, the device dimensions can be further reduced. Computer simulations indicated that the maximum device output voltage, or the number of photodiodes in series, can also be used as a free parameter in order to minimize the device dimensions [27].

C. Contact Material

Injection of high currents through micro size contacts imposes a challenge for the electrode-electrolyte interface. The voltage across the interface is limited by the water window whether the current is injected through faradic or capacitive mechanisms. Several materials have been investigated for their high charge injection capacity (CIC). Iridium oxide (IrOx) contacts obtained either by activation of iridium or sputtering process can achieve a large CIC by anodically biasing the electrode [32]. However, anodic biasing requires additional active circuitry on the device. Therefore, materials that can provide high CIC at the open circuit voltage of the interface may be a better choice for passive microstimulators. Some electrode materials that use capacitive mechanism are tantalum oxide (TaOx), titanium nitride (TiN), and Poly (3,4-ethylenedioxythiophene, PEDOT). In this study, the charge capacity utilized from the contacts was calculated as 436μC/cm² (i.e. 120μA × 0.2ms/5,500μm²), which is well below the charge injection capacity reported for PEDOT (2.3mC/cm²) [33]. This indicates that contacts could be made smaller without exceeding the CIC of the electrode-electrolyte interface. This would leave a larger top surface area for the photodiodes and in turn increase the output current. Thus, the CIC of the contact material directly affects the stimulator output, which can be traded for a smaller device size if necessary [27].

D. Animal Model and the Site of Stimulation

Rats and cats have been used as animal models for ISMS. Anatomical differences are present between the two models, however rat is advocated as a better model for its similarity of upper limb biomechanics to humans [34]. In this study, the forelimb forces were stable and reproducible within each implant. The point of stimulation was “easy” to find in all trials.

Elbow extension was obtained by stimulating the C6 segment at the ventral horn. Various groups have reported that microstimulation of the ventral sites generates extension and the sites that produce flexion are located more dorsally in the cat [35]. Our investigation located a low-threshold area near the C6 ventral horn for generation of elbow extension in the rat spinal cord. Small forces were also observed while stimulating the intermediate and dorsal regions at C5-C7 (data not shown). ISMS was applied in monkeys cervical region to induce forelimb movements [36]. Elbow extension was generated by stimulating C6-C8 segments.

Intraspinal microstimulation of the ventral horn can activate various neuronal compartments; such as motor neurons, interneurons, specific synaptic inputs from different regions, and afferents [37]. Therefore, it is not possible to determine the exact element(s) being stimulated. The site of stimulation was not marked in this study because FLAMES did not include microfluidic channels, nor was it possible to pass DC current without damaging contacts. Therefore, we can only estimate the site of stimulation from the depth of penetration. In the low-threshold area, we suspect that the cathodic contact was located near the alpha motor neuron axons on the ventral side where they bundle before exiting the cord. Axons have lower threshold than soma when cathodic stimulation is used as reported in several studies [38–40]. The force increased as the power increased monotonously in all graphs. The relation between the force and incident power density looks different in each rat, which is possibly due to slightly different implant locations in the spinal cord. Inter-animal neuroanatomical differences may have contributed to the variations in the plots as well.

The largest force recorded in all experiments was 1.08N and the average of the largest forces from fourteen trials was 0.53N. This level of force produced from a rat forelimb is functional considering the body weight of the animals (350–500g). John Stanford et. al. reported a maximum force of 0.46N from rats as they extended their forelimb through a rectangular slot to press on a lever [41].

The threshold current amplitudes needed to stimulate the ventral horn were about 6±2.5μA (mean±std, in low-threshold area) and the largest current used in the high-threshold area was about 120μA, which are comparable to values reported by other groups (Table I) that investigated the feasibility of ISMS to restore locomotion. We did not find a low-threshold area in the interneuronal region, as reported by others (34), in the rat cervical spinal cord (C5-C7) under ketamine/xylazine anesthesia.

Table I.

Comparison of stimulus parameters with other studies.

Current (μA)	Freq. (Hz)	PW (ms)	spinal cord section	Preparation	Ref.
6–120 μA	50	0.2	rat ventral horn in C6	anesthetized/intact	this study
32±19, 1–4*thresh	40–50	0.2	cat lumbosacral cord (multiple regions)	decerebrated/spinalized	[11]
<100	300	0.25–0.3	cat lumbar cord (multiple regions)	anesthetized/spinalized	[42]
1–15	75	0.3	interneuronal region in rat lumbar cord	decerebrated/spinalized	[43]
50–100	40	0.1	cat lumbar spinal cord (multiple regions)	decerebrated/intact spinal cord	[35]

V. CONCLUSIONS

Prototype floating microstimulators were fabricated in a vertical heterojunction structure on a GaAs substrate. Attempts were made to maximize stimulation efficiency by increasing the output voltage with cascaded photodiodes, coating the contacts with a high CIC material, and locating a low-threshold area in the rat spinal cord. The forces generated and in vitro voltage measurements support the premise that sub millimeter size optically activated microstimulators can generate sufficient currents of functional value. Microstimulation of the human spinal cord is predicted to demand higher levels of NIR illumination. The fact that very low levels of laser power could generate functional limb forces in rats is encouraging for translation of this method to the human spinal cord.