A Diagnostic Circuit for Crosstalk Detection in Microelectrode Arrays

Abstract

Current leakage between channels in microelectrode arrays is a sign of device failure and can lead to shorting of neural signals. The purpose of this project is to detect crosstalk between 32 channels of electrodes. We designed an embedded crosstalk detection system that can stimulate each electrode individually with a constant-current pulse and record voltage transients of the stimulated and adjacent electrodes to generate a matrix of crosstalk values. Charge injection in a phosphate buffered saline solution was used to check the condition of each electrode. A semi-wet condition was then used to determine the percent crosstalk between the channels. The analysis showed that there was minimal crosstalk between the electrodes, except for a known physical defect on the probe. The measurement technique enabled by the electronics circuit has the potential to be used in functional testing and screening of implantable devices.

I. Introduction

Microelectrode arrays have been widely used in their applications as neuroprosthetics. Neurostimulation has been successful in treating many neurological disorders including epileptic seizures [1] and spinal cord injuries [2]. The brain-machine interface has also been demonstrated in clinics [3]. However, longevity of the devices is still an outstanding challenge. Crosstalk is undesired coupling between transmission interconnects leading to the electrode sites, and is often a result of insulation failure due to degradation over time [4]. Corrosion, cracking, and bending in the shank can result in electrode performance degradation and lead to device failure [5, 6].

To determine device longevity and diagnose potential failure over time, soak testing at elevated temperatures is often performed by accelerating the effects of temperature and/or relative humidity on the lifetime of devices [7]. It accelerates mechanical failure, which increases the risk of crosstalk in devices. A common method of detection is AC impedance spectroscopy, which been adapted to characterize electrode recording capability over time. An increase or decrease in impedance provides information to identify trace failure and de-insulation of electrode sites [5]. However, it does not adequately detect crosstalk until a gross failure may be present [4].

In this report, a portable embedded crosstalk detection system has been designed to inject a constant-current pulse that is safe for microelectrodes and record the resulting voltage transient signals. The design uses a high-performance microprocessor to detect crosstalk between 32 active electrode sites in an array. The small package size of the printed-circuit board (PCB) allows for field-deploy ability, which is crucial for in vivo experiments and animal studies [8–11]. The compact PCB can be replicated at a low-cost compared to other commercial stimulation instruments. This study focused on the design of the multi-layer PCB and provided validation on the function of the circuit and its potential to detect crosstalk between microelectrodes.

II. Methods

A. Circuit Description

Fig. 1 outlines a multi-layer PCB with an embedded high-performance, 32-bit ARM M7 Cortex microprocessor (STM32F767ZI, ST Microelectronics, Switzerland). The 6-layer, single PCB was designed in Altium Designer electronic design automation software and manufactured by Sunstone Circuit company. The main features of this design include a stimulation pattern generator, voltage transient monitoring, power management, and USB communication with the PC.

Fig 1.

Overview of the PCB design including power parts, microprocessor, biphasic current stimulation system, microelectrode array, voltage waveform monitoring part, and USB to user interface connection.

For the constant-current stimulation, two internal, 12-bit Digital-to-Analog Converters (DACs) of the microprocessor were programmed to output two constant analog step functions. An analog switch and operational amplifier were designed and programmed to generate a biphasic voltage signal. A Howland current pump circuit was used to convert the generated voltage waveform into current for electrode stimulation. The current waveform is used as the input for two 1:16 demultiplexers (DMUXs), which were programmed to switch channels and individually stimulate each electrode of the array. The microelectrode array is connected to the PCB through a 36-pin connector (Omnetics Connector Corp. Minneapolis, MN).

For voltage transient monitoring, two 16:1 multiplexers (MUXs) were programmed to switch between the channels and record the voltage transient waveforms on each electrode in response to constant-current stimulation. The output of the MUXs is directed to an operational amplifier to shift the biphasic signal above zero as the signal must be only positive values for digital conversion. The output of the shifting amplifier is digitized with an internal 12-bit Analog-to-Digital Converter (ADC) of the microprocessor, and sent to the PC.

B. Software

The STM32CubeMX toolchain was used to assign the pins of the microprocessor to certain functions, including USART communication, DACs, ADCs, and GPIO (General Purpose Input/Output) for various other components. Then, a C code was generated in uVision Keil MDK Version 5 IDE, which allowed for coding specific functions in the language of ANSI (American National Standards Institute) C. The programming for this design allows for user input to control the amount of current for stimulation, which is determined by the voltage output of the DAC. In the program, two for-loops were designed to efficiently switch through the channels of electrodes for both stimulation and recording. For instance, if the user wishes to stimulate and record from 10 channels, the 1^st electrode would be stimulated through DMUX control and recorded through MUX control. The MUX would then switch and the voltage transient of the 2^nd electrode would be recorded. This process continues for all of the electrodes. Once the data is collected for all the channels, it can be processed and analyzed in Microsoft Excel.

C. Device for Testing

A customized silicon-based multisite probe with 10 activated iridium oxide film (AIROF) electrode sites, 5 on each shank, was used to test the designed PCB (Fig. 2). The devices have been extensively validated both in vitro and chronic in vivo settings [12–15]. A particular probe was selected for a known defect that occurred during fabrication, resulting in electrode sites 7 and 9 being shorted together. This defect was thought to be beneficial when analyzing crosstalk between the two electrode sites.

Fig 2.

Image of silicon substrate probe with 10 active electrode sites, which channels numbered. The red ovals represent the known shortage between electrode sites 7 and 9 on the bonding pads and electrode sites.

D. Charge Injection

An in vitro experiment in phosphate-buffer solution (PBS, pH 7.4) was conducted to perform a charge injection test on each of the 10 active electrode sites on the device. The purpose of performing a charge injection experiment before taking crosstalk measurements is to validate that the electrode sites are functional. It can be used to detect open connection electrode sites which should be excluded from crosstalk measurements. Voltage transients reveal the amount of current injected that is safe for the microelectrodes without causing damage. A typical current intensity results in a voltage transient of less than ±0.8V, which is generally considered safe for microelectrodes. For this experiment, a large platinum counter electrode was also placed in the solution and was connected to the ground of the PCB to reduce noise. A resistor network comprised of 100 kΩ resistors was used in parallel with the electrodes and connected to ground. The resistor allows for a baseline reference for the access resistance, which gives information on the condition of the electrode site. If the access resistance is greater than 200 kΩ, it indicates failure due to open connections [13]. Each electrode was first injected with a cathodic-first 6 μA biphasic current, with a pulse width of 150 μs and an interphase delay of 30 μs.

E. Crosstalk Measurements

An experiment was performed to determine the amount of crosstalk between each of the 10 active electrode sites. The electrodes were placed in PBS for 30 minutes. They were then removed from the solution, rinsed with DI water and set to dry in air for 5 minutes. This semi-wet condition is to eliminate the PBS as a major current path and to limit potential current leakage to just between the electrode channels. The condition then only allows for possible saline intrusion at the electrode sites and bonding pads that is most likely to have been caused by degradation.

Each channel was stimulated with a 6 μA current. The biphasic current waveform had a pulse width of 150 μs and interphase delay of 30 μs. The voltage transient signals from each electrode were recorded in a serial fashion: the microprocessor of the PCB was set to stimulate the first electrode, record the voltage transient of all 10 electrodes, switch to stimulate the second electrode, record the voltage transient of all 10 electrodes, and continue this process until all 10 electrodes were stimulated and recorded.

III. Results

A. Charge Injection

Fig. 3 shows the voltage transient signals of the 1^st electrode as the amount of current injected was increased. The charge injection experiment performed on all electrodes showed that each site was functional and could be included in the subsequent crosstalk measurements.

Fig 3.

Charge injection experiment in PBS to determine the electrode condition for proper inclusion in crosstalk measurements.

B. Crosstalk Matrix

Fig. 4 shows the voltage transient signals of all 10 electrode sites while channel 1 was stimulated with the current pulse. Although the same electrode channel was stimulated with the same current intensity of 6 μA as in the charge injection experiment, the voltage transients here differed (generally greater) because the crosstalk measurements were performed in a semi-wet condition.

Fig 4.

Voltage transient signals of 10 active electrode sites in the semi-wet condition. Channel 1 was stimulated with a constant-current pulse of 6 μA. Each site was recorded to determine crosstalk between the channels of electrodes.

For quantification, the voltage during the initial cathodic drop at a time of 50 μs was analyzed. This measured time point is 1/3 into the cathodic pulses, which was chosen because it takes time for the current to form a pathway between channels. The voltage of each adjacent channel was divided by the voltage of the stimulated channel to determine the crosstalk ratio and further, converted into a percent value. This analysis was performed for each set of data to create a matrix of crosstalk percent values, which is shown in Table 1. When a channel is stimulated, that channel’s crosstalk will be 100% since it produces a reference value. A zero percent value would mean there is no detectable crosstalk. The known shorted defect between channels 7 and 9 is validated in the table. When channel 7 was stimulated and channel 9 was recorded, there was 98.06% crosstalk. When channel 9 was stimulated and channel 7 was recorded, there was 96.9% crosstalk. The results proved that these two channels are shorted and that our method of diagnosing crosstalk successfully detected defects in devices. The remaining channels had very low crosstalk, which was expected for this experiment because the probe analyzed was in good condition and, therefore, had little crosstalk between electrode sites.

TABLE 1.

Crosstalk percent matrix for 10 active electrode sites. High levels of crosstalk are highlighted in red. The known shortage defect between channels 7 and 9 is confirmed through the results.

Stimulated Channel	Recorded Channel
		1	2	3	4	5	6	7	8	9	10
	1	100	0	0.19	0	0	0	0	0.39	0.77	0
	2	0.39	100	0	0.79	2.17	0	0	0	0.2	0
	3	0	0	100	0	0.39	0.39	0	0	1.95	0
	4	0	0	0	100	0.98	0	0.79	0	0.79	0
	5	0	0	0	0	100	0	0	0.2	0.79	0.2
	6	0	0	1.2	0	0	100	0	0.4	0.4	0
	7	0	0	0	0	0	0	100	0	96.9	0
	8	0.2	0.4	0	0	0.4	0.4	0	100	0.4	0.79
	9	0	0	0	0	0	0	98.06	0	100	0
	10	0.2	0.2	0	0	0	0	0	0	0	100

IV. Discussion

We have designed a portable circuit to individually stimulate up to 32 electrodes with a constant-current pulse and record voltage transient waveforms of all 32 channels. The function of this design enabled rapid and accurate detection of crosstalk between channels of microelectrode arrays through the use of a high-performance microprocessor. The compact design allows for portability and replication at a low cost, which is an advantage over commercial systems.

Through performing the charge injection experiment, we were able to validate the circuit’s capability of delivering a constant-current pulse to an electrode. We were also able to validate that each electrode site had similar properties, which is important for crosstalk detection. The semi-wet experiment confirmed that the circuit is able to individually stimulate 10 electrodes while recording the voltage transients of each stimulated electrode as well as the remaining 9 adjacent electrodes. With further data analysis, we were able to create a crosstalk matrix that determined the percent crosstalk between each electrode. While we showed in this study that the semi-wet measurement was feasible in vitro by eliminating the PBS as a major conductive path, a relative crosstalk measurement is also likely to be possible in vivo with the inclusion of a crosstalk components from the tissue resistance in the brain.

The designed PCB has several advantages over traditional impedance measurements. With USB communication and the fast switching between channels, our crosstalk circuit can stimulate and record all 10 channels of the microelectrode array in under four minutes. Traditional AC impedance measurements are more labor intensive and time-consuming as each electrode must be measured individually, all possible combinations must be evaluated in the case of pair-wise evaluation, and inter channel crosstalk needs to be quantified [16]. Impedance spectroscopy also needs to be performed over a range of frequencies to properly evaluate the conditions of the electrodes and tissue impedances, which would increase the amount of time the process takes. In addition, the method requires a potentiostat or galvanostat for precision, which is bulky and expensive, and may require a separate switch and customized software. Another study used stimulus pulses as test waveforms as done in this study, but could only detect short and open circuits through common reference connections [17].

Further testing of crosstalk over a period of time can provide information on device longevity and failure due to current leakage between electrodes. In particular, soak tested probes at elevated temperatures can be used in future experiments to understand device failure mechanism and further validate our method of crosstalk detection.