Portable wireless electrocorticography system with a flexible microelectrodes array for epilepsy treatment

In this paper, we present a portable wireless electrocorticography (ECoG) system. It uses a high resolution 32-channel flexible ECoG electrodes array to collect electrical signals of brain activities and to stimulate the lesions. Electronic circuits are designed for signal acquisition, processing and transmission using Bluetooth Low Energy 4 (LTE4) for wireless communication with cell phone. In-vivo experiments on a rat show that the flexible ECoG system can accurately record electrical signals of brain activities and transmit them to cell phone with a maximal sampling rate of 30 ksampling/s per channel. It demonstrates that the epilepsy lesions can be detected, located and treated through the ECoG system. The wireless ECoG system has low energy consumption and high brain spatial resolution, thus has great prospects for future application.

. Diagram of the ECoG system. The red block is the ECoG electrode device with an array of 32 flexible microelectrodes, which are used to record ECoG signals of a brain or apply stimulation electrical signal to suppress epilepsy; the yellow block is the electronic circuits which acquire and process ECoG signals; the blue block is the microcontroller unit controlling the ECoG system, and communicate with cell phone ad cloud system for data processing etc.
was designed to restrain the coupling of spike signals and to decrease the noises from the power line. As radio frequency circuits are sensitive to Electro Magnetic Interference (EMI), all wires have been designed as short as possible to reduce the wire inductance and associated noises brought in. Furthermore, polyimide(PI) printed circuit is used to realize this flexible wireless ECoG system as shown in Figure S4 in Supplement. The CC2541 and RHD 2132 are soldered on both side of the PI board to save space.
As the computing power of cell phones is limited, we developed a cloud model to process and recognize real meaning of ECoG signals from a brain. Massive distinct ECoG data can be uploaded to cloud system by APP for processing. The cloud system can easily extract potential patterns and hidden information in ECoG signals of a patient with epilepsy and improve the classification algorithm iteratively by analyzing and integrating large amounts of data. The APP can also receive, store data locally, and display it through different channels in real-time or within a specific time frame. The APP in a cellphone can invoke the pre-trained epilepsy detecting model to identify whether the patient is suffering from epilepsy or not. If it detects signals of epilepsy, it can send a treatment instruction to the user.
Flexible microelectrodes array. The ECoG device has an array of flexible microelectrodes, and could fit to wrinkled surface of the cortex conformally 8,9 , thus it has better contact with more neurons [10][11][12] and has better signal to noise ratio compared to traditional ECoG electrode devices with rigid substrates. Figure 3 is a schematic and photo of the microelectrodes, and the flexible and pluggable ECoG device developed for neural signal recording and electrical stimulation. The electrode device has an array of 32 microelectrodes, with a total width of 16.3 mm and a length of 24.8 mm. The width of a microelectrode is 100 μm at the tip, and each microelectrode has an open surface of 50 μm in diameter at the tip for electrical contact. The 32-microelectrodes array could cover most area of a rat brain, including the important subdomains. There are three large holes of 300 μm in diameter (three black dots in the photo of Fig. 3c) in the ECoG electrode device that are for drug injection.
As the microelectrodes of ECoG device are for implant in brain, their biocompatibility, flexibility and reusability etc have been considered in design and fabrication. Figure 3a shows a cross sectional view of a microelectrode of the ECoG device. It is made on a polyimide (PI) film of 60 μm with excellent flexibility and biocompatibility 9,13 . The metal layer consists of Cr/Ag/Cr three layers with excellent flexibility and good conductivity. A contact window is open for electrical contact for each microelectrode. The microelectrodes were connected with a flexible printed circuit (FPC) board connector to form the ECoG electrode device as shown in Fig. 3(d). As it can be seen that the ECoG electrode array is flexible, and can fit well with wrinkled cortex of a brain, which ensures to record distinct neural signals with high SNR.
The impedance of each microelectrode was assessed at 1 kHz in saline with DF-I (IMP-2, Bak Electronics Inc, CN) to see their suitability for the implant. The impedance of all the 32 microelectrodes is below 120 kΩ with an average value of ~22.7 kΩ, much smaller than 600 kΩ required by implantation 14 , thus it can be used for in-vivo experiments. To test its stability and anti-corrosion properties in biofluid, the flexible microelectrodes were immersed in a 5% saline for 5 days. Close inspection and electrical measurement showed that there is no corrosion and impedance change at all.

ECoG system in-vivo testing. The developed ECoG system was assessed with animal experiments.
Craniotomy was conducted on a healthy adult male Sprague-Dawley rat to expose its cortex and the sterile microelectrode array was placed to cover the left primary sensory cortex of the rat brain as shown in Fig. 4a and b. The normal rat brain signal is shown in Fig. 4c with a typical signal amplitude less than 50 μV. The rat was then injected with penicillin (7.6 × 106 μ/kg) to induce epilepsy 15 . The recorded corresponding ECoG signal is shown in Fig. 4d, exhibiting obvious epilepsy spikes with large signal amplitude over 150 μV. The results clearly demonstrated the ability of the ECoG microelectrodes to record brain electrical signals. The rat ECoG signals from the 32 microelectrodes are shown in Figure S5 in Supplement.
A cellphone was used to receive ECoG signals and display the information through APP, wireless communication and cloud system, with one channel result shown in Fig. 5a. Due to the limited graphics function of the cellphone, only some characteristic points of the signal are displayed. Figure 5b and c show the processed ECoG signals of the rat at normal and epilepsy conditions from four microelectrodes which contact the epilepsy lesions. Compared with the normal state (Fig. 5b), the ECoG signal of the rat under epilepsy (Fig. 5c) has spike signal with very large amplitude, typical characteristic of epilepsy. We could use data mining and pattern recognition algorithm to extract potential information through the cloud, however there are many patterns which is difficult to clearly recognized at the moment.
As the computing power of cellphone is limited, we developed a cloud model to process and recognize the real meaning of brain signals shown in Figure S6 in Supplement, and then realize the brain-cellphone interaction, which is under development 16 . The APP in cellphone can invoke the pre-trained epilepsy detecting model to identify whether the patient is suffering from epilepsy. Besides, massive distinct ECoG data can be uploaded to cloud system by APP for processing. The cloud system can easily extract potential patterns in ECoG signals of a patient with epilepsy and improve the classification algorithm iteratively by analyzing and integrating large amounts of data. The APP can also receive, store data locally, display it through different channels in real time or within a specific time frame. If it detects a wave indicating the epilepsy, it can send a treatment instruction to the user.
ECoG system for rat epilepsy location and stimulation. Based on the recording results and signal processing, we can map out the amplitude of ECoG signals over the measured areas of the rat's brain, and locate the exact epilepsy lesions. Figure 6a and b show the mappings of the brain electrical signal amplitudes measured under normal state and epilepsy state. The red area with the highest amplitude is the most active area under epilepsy and could be identified to be the epilepsy lesions.
Nevertheless, it also could stimulate at the epilepsy nidus after location. Indeed, we conducted this experiment to see if it is viable to achieve epilepsy using constant current electrical stimulation which is widely adoped 17 . Figure 7 shows the recorded response of the rat brain when 100 uA constant current electrical stimulation was applied, and the ECoG signal amplitude of epilepsy spikes reaches mV level. The results clearly demonstrated that the developed ECoG system could apply stimulation at epilepsy nidus. Maybe in the future we could develop a stimulation method how to suppress epilepsy.  Table 1 is the comparison of our ECoG system with other wireless systems. Our system has a higher sampling rate, which is important to obtain detailed and accurate information of the waveform of brain signals. Our system has a larger bandwidth which allows collecting more signals and information. Other advantages include high transmission rate and low power consumption that ensure better communication for a long time.
Since electroencephalogram can directly map human actions, consciousness and emotions, after data pre-processing, feature extraction and pattern recognition, ECoG data can be used in clinical medicine for the treatment of anxiety, insomnia, Alzheimer's disease, brain tumors, epilepsy and other diseases 23 . Furthermore, large amounts of ECoG data are very useful for brain science and neuron science such as functional cognition, brain wave-based control and human-computer interaction, etc. The combination of computer science and neuroscience promote scientists to solve more and more problems about human life such as how to use human brain (i.e. consciousness) to control their behavior, to control machines or even people 24 . Even more amazing is that the brain wave control could be from a cloud computing service, which makes the Internet be the extension of people's brain. Our cell phone based wireless ECoG system has clearly demonstrated the capability of recording 32-channel neuron signals, and interacting with cloud system through APP of a cell phone, and performing electrical stimulation to treat epilepsy. The ECoG system has the capability for wireless communication, powerful signal processing and pattern recognition, yet the device is very small and portable, thus it has great potential for the above mentioned applications.
In summary, we have developed a new wireless brain-cell phone interaction ECoG system to record electrical signal of brain activities. The flexible ECoG microelectrode array consists of 32 channels which can fit conformally on wrinkle structure of cortex with high spatial resolution. Corresponding electronic circuits have also been developed for signal acquisition, processing and wireless communication with cell phone. The in-vivo experiments on a rate have clearly shown the flexible ECoG system can record, process and transmit electrical signals of brain activities to cell phone with good SNR and signal integrity, and demonstrated epilepsy by electrical stimulation through the ECoG microelectrodes. Compared with others, our system has many advantages, including higher sampling rate, larger bandwidth, lower power consumption, high transmission speed and long communication distance, thus demonstrated its great potential for future applications.   Mapping of ECoG signal amplitudes on the rat brain. Mappings of ECoG signal amplitudes from a rat brain under normal state (a) and epilepsy state (b). The red area has the highest amplitude, i.e. most active under epilepsy, which allows us to identify the epilepsy lesions in the brain. Fabrication of flexible microelectrode array. The electrode device has an array of 32 microelectrodes, with a total width of 16.3 mm and length of 24.8 mm made on a PI film. The fabrication process is as follows: A (Polymethyl methacrylate) (PMMA) layer was spun coated on a glass substrate as the sacrificial layer for removal from the glass which was used as a support for easy fabrication, and was baked at 180 °C for 30 min; then a 60 μm PI layer was coated on top and baked at 230 °C for 180 min. After baking, microelectrodes were formed by photolithography and lift-off process. A Cr layer of 20 nm thickness, Ag of 600 nm and Cr of 20 nm were deposited by sputtering in sequence. Both the bottom and top Cr layers were used to improve adhesion of metal with the PI layers. A 600 nm thickness of Ag layer was used to achieve low impedance for better signal acquisition with low noise, yet to have sufficient flexibility. Then, another PI layer about 10 μm was coated on top of the Cr layer for insulation and baked at 230 °C for 180 min. The microelectrode array were patterned and etched by oxygen plasma to open windows of metal layer for electrical contact and holes for drug injection. Au of 60 nm thickness was then electroplated on the windows and holes of the Cr top layer to achieve better contact with neurons and to improve the conductivity of the microelectrodes. Once completed, through acetone steeping, the ECoG electrode with the PI layer was removed from the glass substrate using the PMMA as the sacrificial layer. A flexible printed circuit (FPC) board connector was used to connect the 32 microelectrodes of the ECoG electrode device.
Development of electronic circuits. Electronic circuits were developed for signal acquisition, processing and wireless communication with cell phone. Lower power consumption BLE was used for wireless communication and powered with a button-type battery. A TI CC2541 BLE chip was used as the microcontroller unit (MCU) to control the function of RHD2132 (Digital Electrophysiology interface chips). Details of the circuit chip can be found from Figure S3 in Supplement. Balun circuit was used to convert differential signals to single-ended signals and for impedance matching. The decoupling capacitors were applied to filter spikes in signals, and the shunt capacitors to filter high-frequeData and materials availabilityncy noises. A 32 MHz crystal oscillator was used for normal operation, while the 32 kHz crystal oscillator for sleep mode; An Invert-F antenna with a central frequency of 2.4 GHz and 50 Ω impedance was designed for wireless communication. The designed wireless transmission distance is 10 m for this work, and can be easily upgraded according to applications.