Selective activation of cortical columns using multichannel magnetic stimulation with a bent flat microwire array.

Abstract

Objective:

Cortical neural prostheses that aim to restore useful vision, hearing, and tactile sensations require the ability to selectively target different cortical regions simultaneously. Electrical stimulation via intracortical electrodes has been used to create spatial patterns of cortical activation. However, their efficacy remains limited due to the inability of conventional electrodes to confine activation to specific cortical regions around each electrode. Magnetic stimulation from single bent wires can selectively activate pyramidal neurons while avoiding passing axons, thereby confining activation to small cortical regions. This paper presents a novel bent flat microwire array and demonstrates its effectiveness for selective activation of cortical columns in mouse brain slices.

Methods:

A computational model was developed to compare the spatial resolution of magnetic stimulation from bent wire arrays with 280 and 530 μm tip spacings. The same array designs were fabricated for use in electrophysiological experiments, i.e., calcium imaging (GCaMP6s) of mouse brain slices.

Results:

All fabricated array designs reliably produced spatially discrete cortical activations at low stimulus amplitudes, but the 280-μm-spacing produced strong interference (constructive or destructive) at high stimulus amplitudes, thereby resulting in single strong activations or two asymmetric activations. 4-channel bent wire arrays with spacing of 340 μm avoided the interference and produced clearer spatial patterns of activation than electrodes.

Conclusion:

Bent wire array designs can influence the strength or the spatial resolution of multichannel magnetic stimulation.

Significance:

These results suggest that bent microwire arrays can enhance the selectivity of multichannel stimulation of brain and therefore may help to develop reliable and effective cortical neural prostheses.

I. Introduction

Simultaneous selective activation of multiple targeted regions in the neocortex is critical to the cortical neural prostheses that aim to restore vision [1, 2], hearing [3], and tactile sensations [4]. Multichannel electrical stimulation via microelectrode array implanted in the cortex has been proposed as a means to target different cortical regions (i.e., cortical columns [5]) and much previous work has demonstrated the viability of such stimulation for activation of cortex in animal experiments [3, 6] and in clinical testing [1, 4]. Despite the potential and much ongoing effort, efficacy of the electrical stimulation remains limited, in part due to the inability to precisely target specific types of cortical neurons or confine activation to specific cortical regions around each electrode. Individual electrodes of conventional electrode arrays produce spatially symmetric electric fields, and therefore result in activation of both vertically-oriented pyramidal neurons (PNs) in the local region around the electrode as well as horizontally-oriented passing axons that arise from neurons located in distant cortical regions. The unintended activation of passing axons leads to the spread of activation into the region around the adjacent or distant electrode, and thus diminishes the ability to create precise spatial patterns of neural activity [7–9]. Another limitation arises from the concerns about the long-term viability of the intracortical electrodes during chronic implantation. For example, implantation of electrodes into cortex induces the complex brain tissue responses [6, 10, 11] that lead to the formation of glial scarring around individual electrodes and can increase electrode impedance and stimulation thresholds over time. There have been other approaches utilizing non-penetrating electrodes, e.g., cortical surface electrodes [9, 12], and they can alleviate some of these concerns but requires much higher levels of stimulus current and produces non-focal activation thereby limiting the spatial resolution of multichannel stimulation.

Magnetic stimulation from cortically-implantable bent microwires, previously referred to as microcoils [13, 14], has the potential to overcome many of limitations described above. For example, similar to the submillimeter-sized inductors used in previous studies [15–17], the electric fields induced by the bent wires are spatially asymmetric and can be used to selectively target vertically-oriented PNs while avoiding horizontally-oriented passing axons, thereby confining activation to the local region around the bent wire. The previous study has shown that single bent wires could better confine activation to a small region around the bent wire than the conventional single electrodes [14], and suggested that the bent wires can enhance the spatial resolution of cortical stimulation. In addition, magnetic stimulation uses magnetic fields with high permeability to biological tissues, and therefore the efficacy will not be diminished by glial encapsulation of cortical implants [10]. Finally, unlike electrodes that require direct contact between metal electrode and neural tissue, the bent wire implants can be fully encapsulated with biocompatible dielectric materials [18–20], thereby greatly enhancing the long-term stability of the implants.

Despite the success in demonstrating the enhanced selectivity of single bent wires [13, 14], it is still unknown whether multiple bent wires located close to each other can selectively target different cortical regions. Previous work with multiple large (10–50 mm in diameter) [21, 22] or small (100–500 μm in diameter) [23, 24] coils for non-invasive magnetic stimulation reported concerns about the effect of magnetic coupling (i.e., crosstalk) between two adjacent coils, which may reduce the effectiveness of stimulation. For multichannel bent wire arrays to be useful, it is important to determine whether the spacing between two adjacent bent wires influences the strength of the gradient of the electric field (the driving force for activation [25]) as well as the spatial resolution. Recent advances in the optical neural imaging techniques [8, 26, 27] and the high-definition image sensors (e.g., 4K-pixel-resolution) allow the spatial extent of cortical activation produced by such bent wire array designs to be precisely measured.

Here, to explore the viability of multichannel bent microwire arrays to selectively target different cortical regions, I developed a novel bent wire array design based on Pt-Ir flat ribbon wires and polyimide (PI) tubes. I first built a computational modeling to estimate the effect of the spacing between two adjacent bent wires on the field gradients arising from each bent wire. This model suggests that both the spacing and the stimulus strength influence the level of interference (i.e., constructive or destructive) between the bent wires, thereby influencing the spatial extent of activation as well as the spatial resolution of stimulation. I fabricated the bent wire array designs and evaluated their performance via in vitro calcium fluorescence imaging (GCaMP6s) in mouse brain slices. The calcium imaging results confirmed the modeling predictions that both the array spacing and the stimulus amplitude affect the spatial extent of activation and the spatial resolution. Based on the results, I fabricated a 4-channel bent wire array with optimized spacing, and showed that the 4-ch array allowed for the robust, simultaneous selective activations of multiple cortical columns without causing destructive interference. This design could also produce spatially discrete activations in mouse hippocampus slices. These results strongly suggest that the new bent wire array designs can provide high spatial selectivity for multichannel stimulation of brain and therefore they are highly attractive for use in a variety of neural prostheses and brain-computer interfaces.

II. Materials and Methods

A. Computational modeling of bent flat microwire arrays.

A finite element method (FEM) simulation was performed to calculate the spatial gradient of induced electric fields (E-field) arising from the flow of electric current through the bent microwire array. Similar to the previous studies [13–16], I built 3D models of bent wire arrays using an ANSYS Maxwell 3D (ANSYS, Inc., Canonsburg, PA, USA). Two single-loop Pt flat wires were modeled and placed side by side with a tip spacing of 280 μm (Fig. 1C, ‘Bent wire array A’). Each wire has ‘U’ shape at the tips (top and bottom); the flat wire had a cross-sectional dimension of 10 × 50 μm. The dimension of each loop was 120 × 1000 μm. The area surrounding the array (4 × 4 × 4 mm) was modeled as a homogeneous medium with properties of air. More details of the boundary conditions and other simulation parameters are presented in Section S1 in the supplementary materials.

Fig. 1.

Design and simulation of bent flat microwire arrays for intracortical magnetic stimulation. (A) Schematic diagram of a bent flat microwire. (B) Conceptual diagram of multichannel magnetic stimulation of cortex with bent flat microwire arrays. (C) Model bent wire array A with a tip spacing of of 280 μm. (D) Model bent wire array B with a tip spacing of 530 μm. (E–H) Spatial gradients of the E-fields for the different bent wire array designs in (C) and (D). Top panels show the electric currents flowing in the wires (black arrows) and the resulting induced currents around the bent wires (red arrows). Middle and bottom panels show the distributions of the field gradients along the x-axis (dEx/dx) and the y-axis (dEy/dy), respectively, for the different bent wire array designs and the wire current directions. (I–L) Field gradient profiles along the horizontal plot line (black dashed line in (E)) for the different bent wire array designs and the current directions. Blue and red traces show the dEx/dx and the dEy/dy, respectively. Note that the sign is inverted for the y-axis. The thick gray horizontal dashed lines represent the relative threshold level for a situation in which the wire current is equal to the threshold current amplitude, referred to as 1T. Black horizontal solid, dashed, and dashed-dotted lines represent the relative threshold levels when the wire current amplitude becomes 2T, 4T and 6T, respectively. (M–P) Contour lines indicate the areas for which the dEx/dx (blue) and dEy/dy (red) are respectively suprathreshold. Solid lines represent the wire current of 2T; dashed and dashed-dotted lines represent the wire currents of 4T and 6T, respectively.

I used an eddy current solution mode of the ANSYS and the current input was the sinusoidal current with the frequency of 1 kHz and the amplitude of 1 mA. Based on the solutions from Maxwell’s equations which were described previously [13, 15, 16] the E-field induced by the bent wire array was calculated by the software (Fig. S1 in the supplementary materials). Spatial gradients of the resulting E-field (i.e., dEx/dx and dEy/dy) were then calculated with MATLAB software over the two-dimensional (2D) plane (2 × 1 mm) that was positioned 2 μm above the surface of the bent wire array (Fig. 1, C and E). The same simulation was conducted for other array designs (e.g., ‘Bent wire array B’ in Fig. 1D) to compare the spatial distributions of the field gradients for the different array designs. Similar to previous studies [22, 23], I calculated magnetic coupling for the pair of bent wires with the tip spacing of 280 μm, the shortest spacing used in this study (see Section S2 in the supplementary materials). The results showed very low values for both the coupling coefficient (0.00098) and the mutual inductance (600 fH), indicating very weak magnetic coupling between the bent wires.

B. Fabrication and testing of bent flat microwire arrays.

The bent flat microwire fabrication process was a modified version of the one used in the previous study [13]. While the previous fabrication process was based on a Pt round wire and a single PI tube to fabricate one bent wire, the new process used a Pt flat ribbon wire and a pair of PI tubes. The details of the fabrication process are presented in Section S3 in the supplementary materials. The resulting bent wire structures (Fig. 1A) were fixed with the biocompatible epoxy (MED-301), and then insulated with a 2-μm-thick parylene-C coating. The empty space surrounded by the bent wire was filled with the transparent biocompatible epoxy (MED-301). The total length of the fabricated bent wire array was 25 mm; and the tip length allowed for cortical penetration was 1 mm. Each bent wire channel had a DC resistance of ~20 Ω.

Each bent wire array assembly was tested both before and after each experiment to ensure that there was no leakage of electrical current from the bent wire into the brain tissue (see Section S11 in the supplementary materials). I also monitored the temperature changes on the surface of bent wires during repetitive magnetic stimulation and observed increases of less than 0.5°C, well below the threshold for thermal activation of neurons [28, 29] (see the Section S11 and Fig. S8).

C. Micro-magnetic stimulation drive.

Similar to the previous studies [13–17], the micro-magnetic stimulation drive system was based on the combination of a signal generator and an audio amplifier. The details of the system configuration and the stimulus waveforms are presented in Section S4 in the supplementary materials.

D. In vitro brain slice experiments and calcium fluorescence imaging.

Similar to the previous studies [13, 14], physiological recordings and calcium fluorescence imaging were performed using brain slices prepared from 17–30 days old transgenic mice (Thy1-GCaMP6s; Jackson Laboratory, Bar Harbor, ME). PNs from these animals express a calcium indicator that increases its level of fluorescence when the concentration of intracellular calcium increases, e.g., spiking [30, 31]. Total 27 brain slices (13 mice) were used in this study (2-ch bent wire arrays: 18 slices; 4-ch bent wire array vs. 4-ch electrode array: 9 slices). The care and use of animals followed all federal and institutional guidelines, and the Institutional Animal Care and Use Committees (IACUC) of the Massachusetts General Hospital (IACUC protocol number: 2018N000021; Date of approval: 3/6/2018). Details of the brain slice preparation are presented in Section S5 in the supplementary materials.

Prior to testing bent wire arrays, similar to the previous work [13], the selectivity of magnetic stimulation from a single bent wire to the vertically-oriented PNs was assessed by measuring the responses of individual L5 PNs to magnetic stimulation (more details are presented in Section S6 and Fig. S3 in the supplementary materials). For evaluation of the performance of bent wire arrays, a series of calcium fluorescence imaging experiments were performed; the details of the imaging system and the experimental procedure are presented in Section S7 in the supplementary materials.

E. Data Analysis.

In all statistical analyses unpaired t-tests were used to assess whether the difference between the average values for different stimulation conditions was significant. Differences associated with P values <0.05 were regarded as statistically significant. Variances are reported as standard deviation, ±S.D., or standard error, ±S.E. The details of calculation of threshold levels are presented in Section S8 in the supplementary materials.

III. Results

A. Design and simulation of bent flat microwire arrays.

I developed a novel bent microwire array using Pt-Ir flat ribbon wires and PI tubes (Materials and Methods); the cross-sectional area of the flat wire was 10 × 50 μm. Folding the flat ribbon wires around a thin wall (38-μm-thick) between the two PI tubes allowed me to readily construct multiple consistent sharp bends of microwire without using a series of complex microfabrication processes that require bulky silicon substrates (Fig. 1A). In addition, the cross-sectional area of the flat wire was 25x larger than that of the microfabricated gold wire (2 × 10 μm) on a silicon substrate [13, 14], thereby allowing stronger currents. I designed a 2D multi-shank bent wire array to target different cortical columns independently (Fig. 1B). The cross-sectional area of the final structure was 124 × 54 μm, similar to that of neural probes routinely used in chronic implantation [32], suggesting that it can be safely implanted into cortex without causing significant damage to the surrounding neural tissue [11, 32].

To understand whether the spacing between bent wires influences the spatial resolution of magnetic stimulation, I developed a computational model (Materials and Methods) to estimate the resulting electric fields (E-fields) as well as the spatial gradients of the fields that are induced from the flow of current through the bent wire arrays. Because the field gradient along the length of a neuron or axon is the driving force for activation [25, 33], I was interested in creating strong field gradients along the vertically-oriented PNs in targeted cortical columns (Fig. 1B, thick upward and downward arrows) without simultaneously creating strong gradients along the horizontally-oriented passing axons.

The first design I evaluated was a bent wire array that consisted of two Pt-Ir bent flat wires with a tip spacing of 280 μm (Fig. 1C, ‘Bent wire array A’). The tip spacing was selected based on the previous study [14] in which the spatial extent of cortical activation from a single bent wire was ~300 μm. It is also narrower than the diameter of cortical columns which is approximately 300–600 μm [5]. The stimulus waveform (Materials and Methods) was delivered in the counterclockwise (CCW) and clockwise (CW) directions for left and right bent wires, respectively (Fig. 1E, top panel, black arrows). Such different current directions were used to produce ‘constructive interference’ between the two wires, thereby resulting in overlapping induced currents in the narrow space between the bent wires (Fig. 1E, top, red arrows).

Consistent with the simulations of previous work [13, 14], the resulting E-fields (Ex and Ey) were strongest in the immediate vicinity of the wire (Fig. S1A), and the field gradients (dEx/dx and dEy/dy) were strongest at the corners (sharp bends) of each bent wire (Fig. 1E, middle and bottom). Fig. 1I shows the spatial profiles of dEx/dx vs. dEy/dy along the horizontal ‘Plot line’ passing through the bottom of the two bent wires (Fig. 1E, see the black horizontal dashed line in the middle panel). The sign was inverted for the y-axis to better evaluate the negative peak strength which represents the level of depolarizing current for activation. While the negative peak of dEx/dx were spatially isolated from each other, the two negative peaks of dEy/dy were closely positioned within a distance of 160 μm, thereby resulting in a superimposed field gradient (Fig. 1I, dEy/dy ≈ 1.8 kV/m² at x=0 mm); its strength was approximately 3x lower than that of the negative peak dEy/dy (5.7 kV/m²). Although small, the superimposed field gradient was nevertheless encouraging because it suggests that the constructive interference between the two bent wires might lead to enhancement of the spatial extent of activation of vertical PNs. Such a superimposed field gradient could also be created for the dEx/dx when the polarity is inverted. This would increase the spatial extent of activation of horizontal passing axons. However, unlike the dEy/dy, which showed only one peak along the vertical direction (Fig. 1E, bottom, vertical black and white arrows), the dEx/dx showed multiple peaks with different polarities (Fig. 1E, middle, horizontal black and white arrows) within a short distance (< 400 μm). Such a rapid change in polarity was found less effective for activation in the previous study that used a similar design, e.g., ‘W’-shaped bent wire [14]. Thus, this raises the possibility that the simultaneous stimulation from the bent wire array might be inherently effective for activation of the vertical PNs but not the horizontal passing axons.

Because actual threshold values for activation of cortical PNs have not yet been determined, I estimated the spatial extent of activation from the field gradient curves in Fig. 1I by drawing an arbitrary threshold level that change its vertical position as the stimulus amplitude changes (Fig. 1I, see the four horizontal lines). For example, the thick gray horizontal dashed line indicates a situation in which the stimulus amplitude is equal to the threshold current (referred to as 1T); thin black solid line indicates that the stimulus amplitude becomes 2x of the threshold current (2T). In this manner, I could measure the horizontal extent of suprathreshold field gradient for the different stimulus amplitudes (Fig. S2A), and also construct 2D contour plots that showed the areas for which the dEx/dx and dEy/dy are respectively suprathreshold (Fig. 1M). As expected, the two suprathreshold areas of the dEx/dx (Fig. 1M, blue contour lines) were spatially isolated for all stimulus amplitudes tested (2–6T), whereas the areas of the dEy/dy (Fig. 1M, red contour lines) were merged when the amplitude increased from 2T to 4T. Thus, these results showed that simultaneous stimulation via the bent wires with the narrow tip spacing (280 μm) could produce a strong constructive interference when the stimulus amplitude is higher than 3T, thereby enhancing the spatial extent of activation of vertical PNs.

I next tested whether a situation of ‘destructive interference’ influences the strengths of field gradients from the same array design (280-μm-spacing) and the spatial extent of activation. The current direction of the right bent wire was reversed (CW -> CCW), and therefore the induced currents from each bent wire then flow in the opposite direction between the two bent wires (Fig. 1F, red arrows). As a result, there was no superimposed dEy/dy between the two bent wires (Fig. 1J), and thus, the negative peaks of dEy/dy were spatially separated for all stimulus amplitudes. Figure 1N revealed that the suprathreshold areas of dEx/dx and dEy/dy were spatially separated for the all stimulus amplitudes (2–6T), and slightly smaller than those of the constructive interference (Fig. 1M). The results suggest that the destructive interference between the bent wires with the narrow tip spacing (280 μm) slightly reduces the spatial extent of activation but instead can improve the spatial resolution of stimulation by creating two spatially discrete activations within the small area (~300 μm).

To understand how increasing the tip spacing influences the spatial resolution of magnetic stimulation, I modeled a bent wire array with a tip spacing of 530 μm (Fig. 1D, ‘Bent wire array B’). As expected, the E-fields (Fig. S1, G and H) and the field gradients (Fig. 1, K and L) were not affected by the wire current direction. Therefore, the 2D spatial extent plot (Fig. 1, O and P) showed that the simultaneous stimulation via the bent wires with the large tip spacing (530 μm) produces two spatially discrete activations for all stimulus amplitudes (2–6T). Thus, the modeling results suggest that: (1) the use of constructive interference between the two closely positioned bent wires enhances the spatial extent of activation; (2) the use of destructive interference between such bent wires improves the spatial resolution of activation.

B. Fabrication of bent wire arrays and in vitro experiments.

To compare the actual performance of different bent wire array designs, I fabricated the ‘Bent wire array A’ with a tip spacing of 280 μm (Fig. 2A) as well as the ‘Bent wire array B’ with a tip spacing of 530 μm (Fig. 2B) (Materials and Methods). I conducted a series of calcium fluorescence imaging experiments (GCaMP6s) (Materials and Methods). A 500 μm-thick coronal brain slice that contained V1 was mounted in a recording chamber under an epi-fluorescence microscope with a CMOS camera (Materials and Methods) and the fluorescent response across the slice (1413 × 1035 μm) was evaluated for different bent wire arrays. A fabricated ‘Bent wire array A’ was mounted on a micro-manipulator and the tips were positioned over the brain slice such that the long-axis of each bent wire was parallel to the long-axis of the targeted PN (Fig. 2C). Each tip was then positioned at the approximate boundary between cortical layer 2/3 (L2/3) and layer 4 (L4) so that the strong peak field gradient arising from the tip was delivered directly to the proximal axon of the targeted L2/3 PNs (Fig. 2C), the portion of the neuron known to have the highest sensitivity to stimulation [34]. The results below are based on experiments in 27 slices (13 mice).

Fig. 2.

Bent wire spacing influences the spatial resolution of cortical activation. (A) Schematic diagram of the fabricated bent wire array A with a tip spacing of 280 μm. (B) Similar to (A), but for the bent wire array B with a tip spacing of 530 μm. (C) Epi-fluorescence microscope photograph of a V1 coronal slice from a Thy1-GCaMP6s transgenic mouse; the bent wire array A is positioned directly on the slice and a schematic representation is overlaid to clearly show its location. (D) Similar to (C), except the bent wire array B is positioned in the V1 slice. (E) The peak fluorescence change in response to a simultaneous magnetic stimulation from the bent wire array A. Top panel shows the peak fluorescence change under control conditions (i.e., no stimulation); and bottom panel shows the peak change in response to the stimulation with the current amplitude of 2T (143 mA, 1 kHz half sinusoid, 100 pulses at 100 Hz). Red arrows indicate the wire current direction in each bent wire. The scale bar represents 200 μm. (F) Similar to (E), but with the bent wire array B. (G) The average changes in fluorescence of the L5PN somas in response to magnetic stimulation from the bent wire array A. Blue and red traces show the average fluorescence changes in the cells below Ch1 and Ch2 wires, respectively (white dotted circles in e; n=5 cells for Ch1; n=5 cells for Ch2). (H) Similar to (G), but with the bent wire array B. (I) Similar to (E), but for stronger stimulations. Top and bottom panels show the peak fluorescence changes in response to 4T (287 mA) and 6T (430 mA) stimulations, respectively. (J) Similar to (I), but with the bent wire array B. Note that responses below the Ch1 wire was slightly stronger than that below the Ch2 wire, but both became stronger as the stimulus amplitude increased to 4T and 6T. (K) Spatial profiles of the peak fluorescence changes along the horizontal region of interest (ROI, white dotted rectangular box in (I)) in response to the stimulation from the bent wire array A. Blue, red and green traces represent the averaged peak fluorescence changes (n=3) with stimulation amplitudes of 2T, 4T and 6T, respectively. (L) Similar to (K), but with the bent wire array B.

To determine whether simultaneous magnetic stimulation from the bent wire array A can activate PNs, I delivered a train of stimulus current waveforms to both bent wires (Materials and Methods). The stimulus waveform was delivered in the CW direction for both left (Ch1) and right (Ch2) bent wires (Fig. 2E, bottom, red arrows). The same current direction in both bent wires was chosen to test whether the destructive interference between the bent wires can help to create two spatially discrete cortical activations around each tip. The stimulus amplitude was set to a level 2x higher than the activation thresholds (I_th = 72 ± 8 mA, n=5 cells) measured in preliminary electrophysiological experiments (i.e., patch clamp recording of individual PNs) (referred to as 2T) (Fig. S6H). Consistent with the modeling results (Fig. 1, F and N), the simultaneous stimulation strongly activated a small number of L5 PNs located below each tip (Fig. 2E, bottom, white arrows); and the activated L5 PNs were confined to regions that are spatially separated from each other (Fig. 2E, bottom, white dotted circles). I measured the fluorescence responses of individual L5 PNs below each tip (i.e., cells inside each white dotted circle; n=5 cells for Ch1; n=5 cells for Ch2), and found that the PNs below both tips exhibited robust increases in fluorescence (ΔF/F₀ > 20%) (Fig. 2G, blue and red traces). These results suggest that the bent wire array A could produce two spatially discrete cortical activations within a small cortical region (< 300 μm).

I further increased the stimulus amplitude to explore whether the stronger stimulus could still produce the two spatially discrete activations as the modeling results predicted (Fig. 1N). When the stimulus amplitude increased to 4T (287 mA), the activation below the Ch1 became stronger and wider than that below the Ch2 (Fig. 2I, top). For the stimulus amplitude of 6T (430 mA), which was the maximum strength used in this study, such a difference between the Ch1 and Ch2 activations was further maximized, resulting in a single strong activation below the Ch1 (Fig. 2I, bottom). I plotted the spatial profiles of the peak fluorescence changes (Fig. 2K) along the horizontal region of interest (e.g., ROI, white dotted rectangular box in top panel in Fig. 2I) for the different stimulus amplitudes (n=3 presentations), and found that the activation below the Ch1 became stronger as the stimulus amplitude increased whereas the activation below the Ch2 weakened and eventually disappeared. These results suggest that the bent wire array A could produce a strong destructive interference between the two targeted cortical regions when the stimulus amplitude was higher than 4T.

I next explored whether a larger spacing between the bent wires can avoid the destructive interference that arose for the strong stimulus amplitudes (4–6T). I replaced the ‘Bent wire array A’ with a ‘Bent wire array B’ so that the analogous responses to magnetic stimulation could be measured for the larger tip spacing of 530 μm (Fig. 2, B and D). When the stimulus amplitude was 2T (143 mA), the bent wire array B could also produce, as expected, two spatially discrete cortical activations below each tip (Fig. 2F, bottom, white arrows). Interestingly the two activations were slightly larger but weaker than those from the bent wire array A. Individual neurons below each bent wire tip also showed slightly lower increases in fluorescence (ΔF/F₀ > 10%) (Fig. 2H, see blue and red traces). Such a weak but wide activation observed here likely arose because the larger tip spacing may reduce the level of interaction between the two activation areas, such as excitatory and inhibitory synaptic interactions, although I did not attempt to identify the underlying mechanisms.

When the stimulus amplitude increased to 4T (287 mA), the bent wire array B could still produce two spatially discrete cortical activations (Fig. 2J, top, black arrows). For the strongest stimulus amplitude of 6T (430 mA), I did not observe any significant interactions between the two activations (Fig. 2J, bottom, black dotted circles). The horizontal profiles of the peak fluorescence changes (Fig. 2L) showed that the bent wires with the larger tip spacing (530 μm) could produce the two strong distinct cortical activations across all stimulus amplitudes tested (2–6T). Thus, these results suggest that the spacing between the bent wires as well as the strength of stimulus both influence the spatial resolution of cortical activation.

C. Using interference between two bent wires to shape the spatial extent of cortical activation.

The finding of the destructive interference arising between the two bent wires with the tip spacing of 280 μm (Fig. 2, I and K) led me to question whether constructive interference between the same array can enhance the spatial extent of cortical activation. To better observe the spatial distribution of activation around the two bent wire tips, I adjusted the imaging area to a smaller area (534 × 400 μm) and thereby enhanced the signal-to-noise ratio (Fig. 3A). To evaluate the ‘constructive interference’ between the bent wires, the stimulus waveform was delivered in the CCW and CW directions for the Ch1 and Ch2, respectively (Fig. 3A, left panel, red arrows). The stimulus amplitude was set to 3T (215 mA) that was the level at which simultaneous stimulation was expected to produce the suprathreshold field gradients between the bent wires (see Fig. 1, I and M). Consistent with the modeling results, the bent wire array A resulted in a strong activation between the two bent wires (Fig. 3A, left panel). I measured the fluorescence changes (ΔF/F₀) in three different ROIs (Fig. 3A, left, the boxes labeled ‘1’, ‘2’, and ‘M’). The ROI ‘M’ (Fig. 3A, right panel, red trace) showed the strongest fluorescence changes (ΔF/F₀ > 120%) during the repetitive stimulation, while the peak ΔF/F₀ values in ROIs ‘1’ (blue trace) and ‘2’ (green trace) were 57% and 12%, respectively. These results suggest that the bent wire array A indeed produced the constructive interference, thereby resulting in a strong activation in the space between the tips.

Fig. 3.

Interference between two closely positioned bent wires shapes the spatial extent of cortical activation. (A) Left panel shows the peak fluorescence change in response to simultaneous magnetic stimulation from the bent wire array A. Red arrows indicate the wire current direction in each bent wire – counterclockwise (CCW) direction for Ch1; clockwise (CW) direction for Ch2. Note that the wire current in each bent wire flows in the opposite direction in order to create an induced current superimposed between the two bent wires (i.e., constructive interference). Right panel shows the changes in fluorescence which were measured at three different ROIs in the left panel. Blue, red and green traces show the responses at the ROIs, ‘1’, ‘M’, and ‘2’, respectively. (B) Similar to (A), but for the wire current flowing in the reverse direction (CW for Ch1; CCW for Ch2). (C) Similar to (A), but for the wire current flowing in the same direction in each bent wire (CCW for Ch1&2). Note that the same wire current direction in each bent wire produces two induced currents flowing in opposite direction between the two bent wires (i.e., destructive interference). (D) Similar to (C), but for the wire currents in the reverse direction (CW for Ch1&2).

To explore whether the polarity of the field gradient influences the spatial distribution of activation, I reversed the current direction of each bent wire (Fig. 3B, left panel, red arrows) and repeated the same measurement. Similar to the modeling predictions, reversing the current directions resulted in reversing the areas of activation, e.g., the location of strong cortical activation moved from the center to the area close to the Ch2 (compare the locations of green colored regions in the left panels of Fig. 3, A and B). As a result, while the peak ΔF/F₀ in ROIs ‘1’ and ‘M’ were both significantly decreased (e.g., ROI ‘1’, blue trace, 120 to 62%; ROI ‘M’, red trace, 57 to 9.5%), the peak ΔF/F₀ in ROI ‘2’ (green trace) showed a strong increase from 12 to 76%. These results suggest that the current direction of each bent wire indeed influences the spatial distribution of field gradient and thus can be used to modulate the spatial distribution of cortical activation.

To directly compare the strengths as well as the spatial extents of cortical activations for the two different interferences (‘constructive’ vs. ‘destructive’), I conducted the same measurements for the destructive interference (Fig. 3, C and D). When the stimulus waveform was delivered in the CCW direction for both the Ch1 and Ch2 (Fig. 3C, left, red arrows), a moderate level of activation (ΔF/F₀ > 67%) appeared below the Ch2 (Fig. 3C, right, green trace), whereas there was no activation below the Ch1 (blue trace). The spatial extent of the activation (Fig. 3C, left panel) was much smaller than those of the constructive interference (Fig. 3, A and B, left panels). Reversing the current direction from CCW to CW (Fig. 3D, left, red arrows) resulted in a similarly moderate activation (ΔF/F₀ > 63.5%) appearing below the Ch1 (Fig. 3D, right, blue trace), whereas a much weaker activation (ΔF/F₀ > 23.5%) below the Ch2 (green trace). The spatial extent of the activation here (Fig. 3D, left panel) was also smaller than those of the constructive interference (Fig. 3, A and B, left panels). Thus, these results raise the possibility that the use of interference between two closely positioned bent wires may be useful for precisely modulating the strengths as well as the spatial extents of activation within a small cortical region. Further details of the comparison between the simulated field gradients and the calcium imaging results are presented in Section S9 and Fig. S4 in the supplementary materials.

D. Selective activation of single cortical columns with a 4-channel bent microwire array.

To selectively activate single cortical columns in different locations of mouse V1, I designed a 4-channel bent microwire array (Fig. 4B) that consisted of four bent wires with a tip spacing of 340 μm. The tip spacing of 340 μm, which was slightly larger than that of 280 μm, was selected to reduce the level of destructive interference between bent wires and thereby improve the spatial resolution of stimulation. I conducted a computational modeling of the 4-ch array (Fig. S5) and found that the simultaneous stimulation from the 4-ch array creates four spatially separated suprathreshold field gradients below each channel (Fig. S5, E and F) for the amplitudes of 2–6T.

Fig. 4.

Selective activation of single cortical columns with bent microwire arrays. (A) Schematic diagram of a 4-channel microelectrode array (500 kΩ, Pt-Ir) with a tip spacing of 340 μm. (B) Similar to (A), except a 4-channel bent microwire array with a tip spacing of 340 μm. (C) Photograph of the microelectrode array positioned in a V1 coronal slice (GCaMP6s). (D) Similar to (C), but for the bent wire array positioned in a V1 slice. (E) The peak fluorescence changes in response to electrical stimulation from each electrode. The stimulus consisted of 100 pulses delivered at a rate of 100 Hz. Individual pulses were biphasic rectangular waveforms that were 200 μs in duration and 30 μA in amplitude; pulses were cathodic-first without an interphase-interval. (F) Similar to (E), but for magnetic stimulation from each bent wire. The stimulus was identical to that used in Fig. 2, E and F (143 mA, 1 kHz half sinusoid, 100 pulses at 100 Hz). (G) The average changes in fluorescence which were measured at four different ROIs (circles below electrode tips) in (E) for the current amplitude of 10–60 μA (ES 0.67–4T) (data from Ch1–4, n=4 slices). (H) Similar to (G), but for the ROIs in (F) for the amplitude of 57–287 mA (MS 0.8–4T) (Ch1–4, n=4 slices). (I) Average peak ΔF/F₀ for ES vs. MS as a function of the stimulus amplitude (Ch1–4, n=4 slices). (J) Averaged horizontal extents of cortical activation for the electrode vs. bent wire (data from Ch2&3, n=4 slices). The horizontal extent was the full width at half maximum (FWHM). Asterisk indicates statistically significant difference (two tailed t-test: * p = 0.0211; ** p = 0.0004). Error bars indicate Standard Error (S.E.).

In order to evaluate the performance of the 4-ch bent wire array, I measured the fluorescent responses to electric stimulation from conventional monopolar electrodes. Since the monopolar electric stimulation can produce spatially symmetric electric fields and thereby activate both the vertical PNs as well as horizontal passing axons (Fig. S6A), it can provide a good control for assessing the selectivity of the bent wire array. Electric stimulation (Materials and Methods) was delivered via a 4-channel microelectrode array (Fig. 4A) that consisted of four conventional electrodes (500 kΩ, Pt-Ir, Materials and Methods) with the same tip spacing (340 μm). Each tip of the electrode array was positioned in the approximate boundary between the L2/3 and L4 (Fig. 4C). The stimulus train consisted of 100 pulses delivered at a rate of 100 Hz. Individual pulses were biphasic rectangular waveforms that were 200 μs in duration; pulses were cathodic-first without an inter-phase-interval. This pattern of stimulation is similar to those previously shown to elicit behavioral responses in vivo in non-human primates (NHPs) [6, 35]. The amplitude was 30 μA, which was a level 2x higher than the activation thresholds (I_th = 15 ± 2 μA, n=5 cells) (i.e., 2T) measured in the electrophysiological experiments (Materials and Methods; Fig. S6G). Note that this current level (i.e., 30 μA) is considered safe and within the range (20–100 μA) commonly used in intracortical stimulation studies with electrodes of 0.5–1 MΩ [6, 35]. The stimulus current was delivered via one of the electrodes; consistent with the previous studies [7, 14, 36, 37], each electrode produced a strong cortical activation (Fig. 4E) but the spatial extent of activation was not only spatially expansive but also highly variable depending on the different electrode locations (compare the horizontal extent of activation for Ch3 and Ch4). The strong activations (ΔF/F_MAP > 80%, see light blue zone) generally spread extensively to the different cortical layers (Fig. 4E, vertical arrows) along the vertical direction as well as to the different cortical columns along the horizontal direction (horizontal arrows). The wide lateral spread was also found with stimulation amplitudes as low as 18 μA (1.2T) (Fig. S6C); this is consistent with the results from the previous studies of electric stimulation using 20 μA [13, 14].The radius of activation (i.e., suprathreshold area) for 2–200 μA current amplitudes was estimated to be 5–447 μm from the electrode tip in previous studies [6, 35]. For the current level used here (i.e., 30 μA), the radius of activation was estimated to be 67 μm, which is much smaller than the results of this study (e.g., ~500 μm) (Fig. 4E, Ch3). This supports the notion that while the conventional electrodes produce spatially-confined field gradients, but they are spatially symmetric and thus unable to avoid the activation of horizontal passing axons (Fig. S6A).

The electrode array was replaced with the 4-ch bent wire array (Fig. 4D) and a 100 Hz train of 100, 1-kHz half sinusoid waveforms at 2T (143 mA) was delivered to one of the bent wires. Similar to previous studies [13, 14], single channel magnetic stimulation produced a spatially-confined cortical activation around each bent wire (Fig. 4F). It should be noted that the imaging results contain some image artifacts (Fig. 4, E and F, see bright thin lines appearing around non-active electrodes & bent wires which were pointed by white arrow heads). They were generated during the image subtraction process, especially when there was a fluctuation of the liquid in the perfusion bath or a small displacement of imaging plane during the calcium imaging. The artifacts were strongest at the boundary between dark metallic parts (electrodes/wires) and transparent insulation coatings (reflective, i.e., shiny) as well as at the uneven surface of epoxy fill (Fig. 4D, see white arrow heads). However, the artifacts did not extend by more than 10 μm (smaller than single PNs) from the implants, and therefore, it was readily distinguished from actual activation and excluded from the image analysis.

The horizontal extents of strong activation (ΔF/F_MAP > 80%) ranged from 174 to 456 μm (white horizontal arrows), which were narrower than the diameter of single cortical columns (300–600 μm [5]). To quantitatively compare the horizontal extent of activation for the electrode vs. bent wire, I measured the FWHM for two channels in the middle of each array, e.g., Ch2 and Ch3, in 4 different brain slices (Fig. 4J). At the lowest suprathreshold amplitude of 1.2T, the average FWHM for the electrode was 174 ± 16 μm (S.E.), and it was 3.2x larger than the level for the bent wire (e.g., 54 ± 3 μm (S.E.)). For the stimulus amplitude of 2T, the FWHM for the electrode was 431 ± 53 μm (S.E.), and it was slightly larger (~1.3x) than that for the bent wire (e.g., 331 ± 39 μm (S.E.)). When the stimulus amplitude further increased to 4T, the average FWHM for the electrode rapidly increased to 811 ± 133 μm (S.E.), whereas the level for the bent wire slightly increased to 448 ± 43 μm (S.E.). The small increase in the FWHM for the bent wire was encouraging because it suggests that each bent wire can confine the strong activation to a single cortical column. I was unable to compare the FWHM for amplitudes higher than 4T because the electrical stimulation at 6T produced a strong activation over the entire imaging area (1413 × 1035 μm). Even for the 6T, the FWHM for the bent wire did not show any significant increase (e.g. 603 ± 26 μm (S.E.)). Thus, the results suggest that the bent wire better confines cortical activation to a small region in cortex than the conventional electrode for both the low (1.2T) (two tailed t-test: **p = 0.0004) as well as the strong stimulus amplitudes (≥4T) (two tailed t-test: * p = 0.0211).

To compare the temporal changes in fluorescent intensity of individual PNs for the electric vs. magnetic stimulation, I measured the fluorescence responses of the L5 PNs below each tip (i.e., cells inside each circle C1–4; n=4 slices) for different current amplitudes (ES: 10–60 μA; MS: 57–287 mA) (Fig. 4, G and H). The electric stimulation (100 pulses at 100 Hz) resulted in a fast increase in fluorescence for the first 250 ms for all stimulus amplitudes (10–60 μA) and then showed a gradual decrease for low and moderate intensities (18–30 μA) or a secondary peak for the strongest level (60 μA); whereas the magnetic stimulation showed a change in fluorescence increasing at different rates depending on intensity (Fig. 4H). While I did not investigate the underlying mechanism in detail, it seems likely that the symmetric electric fields from the electrode may activate both the axons of targeted PNs (i.e., direct activation [25, 34, 38–40]) as well as horizontal synaptic afferent inputs to the PNs (i.e., indirect activation [39–41]), and therefore the fluorescence level increases rapidly regardless of the stimulus intensity; whereas the asymmetric fields from the bent wire can preferentially activate the vertical axons (direct activation) (Fig. S3, C and E) and some vertical synaptic inputs (indirect activation), so the rate of increase in fluorescence can be precisely controlled by the stimulus intensity.

The difference in spatial and temporal responses between electric and magnetic stimulation may arise due to non-linear scaling differences when the two thresholds (i.e., electric and magnetic) are scaled, e.g., multiplying the threshold for electric stimulation by a factor of 1.2 or 2 is somehow not equivalent to multiplying the threshold for magnetic stimulation by the same factor. Therefore I compared the average peak fluorescence increase for electric and magnetic stimulation (Fig. 4I). At 1.2T, the peak increase (ΔF/F₀) for electric stimulation was 2.36x larger than that for magnetic stimulation (ES: 51 ± 24 % (S.E.); MS: 21 ± 12% (S.E.)), but the difference was not statistically significant (two tailed t-test: p = 0.1289). At higher intensities (2–4T), the peak increase for electric stimulation became highly similar to that for magnetic stimulation (e.g., ES 2T: 102 ± 2% vs. MS 2T: 83 ± 15%, p = 0.1023; ES 4T: 184 ± 11% vs. MS 4T: 198 ± 19%, p = 0.3589). These results suggest that the normalization of stimulus intensity based on the two thresholds can provide quantitatively accurate predictions on the peak calcium signals in targeted PNs arising from two different stimulations. This also supports the finding that magnetic stimulation with bent wires can better confine activation to a small region than electric stimulation with conventional electrodes without compromising the strength of stimulation. But, it should be noted that while the use of calcium imaging and burst stimulation (100 Hz) was useful to measure the spatial extent of cortical activation for the two different stimulations, it did not allow discrimination of the direct activation by each stimulation pulse because the calcium imaging results include both the direct and indirect activation (e.g., recurrent and local inputs).

E. Selective activation of multiple targeted cortical columns of V1 with the 4-ch bent wire arrays.

To explore whether the 4-ch bent wire array can selectively activate multiple targeted cortical columns in V1 simultaneously, I measured the fluorescent responses to simultaneous magnetic stimulation delivered via two bent wires with different tip spacings (Fig. 5). Prior to testing the bent wire array, I measured the fluorescent responses to simultaneous electric stimulation (Fig. 5, A and C). The stimulus current (100 Hz train of 100 stimuli) at 4T (60 μA) was delivered via Ch2 and Ch3 electrodes with a tip spacing of 340 μm (Fig. 5A); as expected, each electrode produced a strong cortical activation (Fig. 5A, ΔF/F_MAP > 140%, green zone) around the electrode but the two activations were not spatially separated, thereby resulting in a single, spatially large activation (see white zone, ΔF/F_MAP > 120%, surrounding the Ch2 and Ch3 in Fig. 5A). The activations also extensively spread in horizontal directions, resulting in a significant level of activations (ΔF/F_MAP > 80%, light blue zone and black horizontal arrow) in regions below inactive channels (Ch1 and Ch4). As a result, the ROIs below the two channels showed strong increases in fluorescence (Fig. 5B, Ch2: red trace, ΔF/F₀ > 128%; Ch3: green, ΔF/F₀ > 130%); the ROIs below other channels also showed moderate activations (Fig. 5B, Ch1: blue, ΔF/F₀ > 24%; Ch4: purple, ΔF/F₀ > 39%). When the same electric stimulation was delivered to the Ch1 and Ch4 electrodes with the largest tip spacing of 1020 μm (Fig. 5C), although two separate strong cortical activations appeared around the Ch1 and Ch4 (ΔF/F_MAP > 120%, white zones), the spatial extent of activation from the Ch1 was significantly wide, thus resulting in considerable activations (ΔF/F_MAP > 80%, light blue zone in Fig. 5C) in the entire area of L5 (Fig. 5D; Ch1: blue, ΔF/F₀ > 190%; Ch2: red, ΔF/F₀ > 62%; Ch3: green, ΔF/F₀ > 20%; Ch4: purple, ΔF/F₀ > 130%).

Fig. 5.

Selective activation of multiple targeted cortical columns of V1 with bent microwire arrays. (A) The peak fluorescence changes in response to simultaneous electric stimulation from electrodes Ch2 and Ch3. The tip spacing between the Ch2 and Ch3 was 340 μm. The stimulus amplitude was 4T (i.e., 60 μA). (B) The changes in fluorescence which were measured at four different ROIs (circles below electrode tips) in (A). (C) Similar to (A), but for electric stimulation from electrodes Ch1 and Ch4. The tip spacing between the Ch1 and Ch4 was 1020 μm. (D) Similar to (B), but for ROIs in (C). (E) Similar to (A), but for magnetic stimulation from bent wires Ch2 and Ch3. The tip spacing between the Ch2 and Ch3 was 340 μm. The stimulus amplitude was 4T (i.e., 287 mA). (F) Simiar to (B), but for the ROIs in (E). (G) Similar to (E), but for stimulation from bent wires Ch1 and Ch3. The tip spacing between the Ch1 and Ch3 was 680 μm. (H) Similar to (B), but for the ROIs in (G). (I) Similar to (E), but for stimulation from bent wires Ch1 and Ch4. The tip spacing between the Ch1 and Ch4 was 1020 μm. (J) Similar to (B), but for the ROIs in (I).

Subsequently, the electrode array was replaced with the 4-ch bent wire array, and the stimulus waveform (100 Hz train of 100, 1-kHz half sinusoid waveforms at 4T, i.e., 287 mA) was delivered in the CW direction to the Ch2 and Ch3 bent wires (Fig. 5E, red arrows); the tip spacing between the pair was 340 μm. Interestingly, even for this narrow spacing, the strong (4T) simultaneous stimulation produced two spatially discrete cortical activations in L5 (Fig. 5E, ΔF/F_MAP > 100%, white zones, see horizontal arrows); the ROIs below the two active channels showed strong fluorescent responses (Fig. 5F, Ch2: red, ΔF/F₀ > 91%; Ch3: green, ΔF/F₀ > 81%) whereas the ROIs below other inactive channels did not show any changes (Fig. 5F, blue and purple). This suggests that the 4-ch bent wire array with the narrow tip spacing of 340 μm can activate two adjacent cortical columns simultaneously. When the same stimulation was delivered to other pairs with larger tip spacings (e.g., 680 and 1020 μm), as expected, two discrete cortical activations appeared around the active channels (Fig. 5, G and I); the ROIs below the active channel pairs showed similarly robust increases in fluorescence (Fig. 5, H and J; ΔF/F₀ > 47% for all active channels) whereas those below the inactive channels did not. This suggests that each bent wire can confine activations to a targeted single cortical column in V1 during simultaneous stimulation, thereby allowing for the creation of a clear spatial pattern of cortical activations. This feature may be highly useful for visual prostheses [1, 2, 9, 12] as well as other cortical implants [4] that require enhanced spatial resolution. In additional experiments with mouse hippocampus slices, the 4-ch array also successfully produced spatially discrete activations (see Section S10 and Fig. S7 in the supplementary materials).

IV. Discussion

The new bent wire design based on the flat ribbon wires and the PI tubes (Fig. 1A) has several advantages when compared to previous designs such as microfabricated bent wires on silicon substrates [13, 14] and bent round wires [13]. For example, the use of flat ribbon wires allows for the fabrication of consistent sharp bends of microwire without using bulky substrates (100 × 50 μm). As the large substrate was removed, I could use thick (10 × 50 μm) flat wires while maintaining the similar implant dimension. In addition, compared to the bent round wires based on 25-μm-diameter Pt-Ir wires [13], the flat wire was 2.5x thinner (i.e., 10-μm-thick), and thus allowed sharper bends at the tips as well as larger optically transparent windows in the middle of bent wire structure. The optical window (100 × 1000 μm) that was filled with the transparent epoxy allowed me to monitor cortical activations that arose behind each bent wire with calcium fluorescence imaging methods (Fig. 2, I and J; Fig. 3), suggesting that the new design is advantageous for using various optical imaging methods including voltage sensitive dye imaging.

Physiological experiments showed that at low stimulus amplitudes (2T), bent wire arrays with the narrow and wide tip spacings (280 and 530 μm) both produced the spatially separate cortical activations (Fig. 2, E and F), whereas at higher amplitudes (4 and 6T), the two activated areas remained separated for the wide tip spacing (Fig. 2J) but not for the narrow spacing (Fig. 2I). At the high amplitudes (3–6T), the pair of bent wires with the narrow tip spacing (280 μm) showed strong constructive or destructive interference depending on the current direction in each bent wire. The constructive interference resulted in a strong and spatially wide single cortical activation between the pair (Fig. 3, A), whereas the destructive interference resulted in spatially asymmetric cortical activations (Fig. 3, C and D; Fig. S4, B and C). The strong destructive interference and the resulting asymmetric activations were similar to those of the previous work using ‘W’-shaped bent wire designs with much narrower tip spacings (80 μm) [14], supporting the notion that the interaction between two adjacent field gradients with different polarities can reduce the effectiveness of activation. It is important to note that the levels of destructive interference observed in the physiological experiments (Figs. 2I and 3) were much stronger than those of the modeling results (Fig. 1, M and N; Fig. S4, B and C). This suggests that although the model used in this study can produce qualitatively accurate predictions, further improvements, e.g., the use of realistic neuron models, are needed to produce quantitatively accurate predictions.

Based on the results with two bent wires with different spacings (280 and 530 μm), I fabricated a 4-channel bent wire array with a tip spacing of 340 μm (Fig. 4B), and showed that the 340-μm tip spacing allowed for robust, simultaneous selective activations of multiple targeted cortical columns without causing the destructive interference (Fig. 5). This study also compared the two-point resolution of the bent wire array with that of the conventional electrode array. When the stimulation at the high amplitude (4T) was simultaneously delivered to two different channels, the bent wire array showed 3x finer two-point resolution (~340 μm) than the electrode array (~1020 μm). The resolution of the electrode array (~1020 μm) was similar to those reported in previous studies with intracortical electrode arrays implanted in NHPs (1–1.6 mm) [6] and in human subjects (0.5–1 mm) [1], suggesting that the methods of this study, used to estimate the spatial extent of activation in V1, may provide an accurate estimate of the spatial extent of visual percepts (i.e., phosphene sizes) elicited by stimulation of V1. Thus, the spatial resolution of the bent wire array was approximately ~4x smaller than those of intracortical electrode arrays (~1.6 mm) [1, 6], and also ~10x smaller than those of surface electrode arrays (2–4 mm) [9, 12].

Although these experiments used strong stimulus currents for the bent wires to assess the maximum spread of activation, the minimum current (i.e., threshold current) required to activate cortical PNs was as low as ~72 mA, and thus the actual peak power level calculated for the 1-mm-long bent wire implant was ~3.9 mW (i.e., 72-mA current flows through 0.76-Ω Pt-Ir wire). This level is 1.6–12.5x higher than that for intracortical electrical stimulation (e.g. 0.3–2.3 mW) [1, 6], but is comparable to that for existing clinical devices, e.g. the power levels for deep brain stimulation range from 2–25 mW [42]. For use in the applications such as visual prosthetics that require a large number of channels (e.g., 100 × 100), the power level of the bent wire array should be further minimized to alleviate concerns about the relatively high power consumption. Some preliminary studies (not shown) have shown that the bent wire array consisting of 2x smaller (e.g., 10 × 25 μm) flat wires could reduce the threshold current by half (i.e., 72 to 36 mA), thereby lowering the power level to 1.9 mW, e.g., P_10×25μm = (I_10×50μm/2)² × (2 × R_10×50μm) = P_10×50μm/2, suggesting that the optimizing bent wire designs could further minimize the power consumption. The use of materials that enhance magnetic field strength (e.g. ferrite, Mu-metal, and permalloy) may also reduce power levels even further although concerns about the biocompatibility of these materials will need to be investigated.

The micro-magnetic stimulation drive system was based on the combination of a signal generator and an audio amplifier (i.e., voltage amplifier). The use of the voltage amplifier did not allow for the precise control of the stimulation current level on each channel with varying impedance, and therefore, for use in multichannel magnetic stimulation, it is desirable to use a current amplifier that can provide the desired current even to channels with varying impedances.

V. Conclusion

Here, I showed that a novel bent flat microwire array can selectively activate multiple targeted cortical columns of V1. By using Pt-Ir flat ribbon wires and PI tubes, I designed and fabricated a 2D multi-shank bent microwire array for use in multichannel magnetic stimulation of cortex. Computational simulations predicted that each bent wire can produce strong electric field gradients within a small area around the tip (~140 μm), and that the array design, i.e., the spacing between the bent wires, influences the spatial extent of suprathreshold field gradients as well as the spatial resolution of intracortical magnetic stimulation. Consistent with these findings, a series of electrophysiological experiments revealed that bent wire array can selectively activate different cortical columns simultaneously, and that both the spatial extent of activation and the spatial resolution were influenced by the spacing between the bent wires as well as the strength of stimulus. The precise control of simultaneous activation of targeted cortical columns suggests that the bent flat microwire array has the potentials to advance the spatial resolution of multichannel cortical stimulation for use in a wide variety of neuroscience studies as well as cortical neural prosthetics.