Synopsis

Keywords: New Devices, Neuroscience

We propose a new device for neuroscience studies: the MRDust, a sub-mm wireless programmable neural recording mote with on-device memory and compute. It receives power via focused ultrasound, records neural signals in burst mode, and uses a micro-coil to perturb local magnetic fields to achieve data uplink via dynamic MRI signal modulation. We demonstrate proof-of-concept experiments in which digital information is encoded in images of an SE-EPI dynamic sequence, and in which a piezoelectric harvester can harvest enough ultrasonic power to sustain device operation, and receive control signals through amplitude modulation.

Introduction

Being able to record neural activities at high temporal and spatial resolution with whole-brain coverage allows us to better understand, diagnose and treat neurological disorders, but remains a challenge¹. Non-invasive methods like functional MRI offer information about brain structure/function, but exhibit low temporal resolution. Tethered devices (Utah Array²/Neuropixel³) can record in a small cortex region at high temporal resolution, but have low spatial coverage and are extremely invasive. Wirelessly powered miniaturized implants are an appealing alternative as they cause minimal tissue damage⁴. Unfortunately, conventional wireless power and data link modalities (RF/ultrasound) suffer significant tissue propagation loss, making it extremely difficult to reach deep brain^5,6. Device localization also becomes challenging with an increased number of implants.
In this work, we present a new concept called MRDust, which aims to alleviate some of the aforementioned issues by utilizing focused ultrasound for power and data downlink, and dynamic MR image modulation for data uplink. It has the potential to yield micro-sized devices that record deep brain neuronal activities at multiple sites simultaneously. Devices can be implanted in clusters in an animal brain or a human brain undergoing surgery. Here we demonstrate several proof-of-concept feasibility experiments: we show that digital information can be encoded in dynamic MRI scans by DC current modulation in a micro-coil, and that enough power can be harvested by a piezoceramic to support device operation. We designed an integrated circuit (IC) which performs power harvesting, uplink and downlink communication, which is under fabrication.

Methods

Device Overview:
Each device is envisioned to contain a piezoceramic for power harvesting, a pair of electrodes to measure neural signals, and a programmable chip with compute and storage capabilities to perform neural data recording, processing and uplink communication (Figure 1a). The device operates in the following order in each recording cycle: 1) Focused ultrasound is used to power the devices and send control signals by amplitude modulation (Figure 1b). 2) Devices are signaled to record neural data in burst mode for short intervals, process them and store on-chip (Figure 1c). 3) To perform data uplink, DC current is driven through a micro-coil on-chip to modulate the local magnetic field, encoding stored data frame-by-frame by modulating voxel intensity (Figure 1d). We hypothesize that modulating the MR signal using DC currents in a micro-coil is more efficient than RF/Ultrasound communication from a micro-device.
Pulse Sequence:
Data transmission is achieved by syncing the coil current to a dynamic pulse sequence (GRE-EPI/SE-EPI) via ultrasonic downlink. By passing a DC current through the micro-coil, we generate a local magnetic field to dephase surrounding spins, resulting in diminished voxel intensity. During GRE-EPI, coil current is turned on or off throughout the TR to transmit a digital “1” or “0” respectively (Figure 2a). During SE-EPI, the coil’s current is reversed after the spin-echo to suppress refocusing of local spins when transmitting a “1”, and turned off when transmitting a “0” (Figure 2b). In both cases, digital “1” exhibits a lower voxel intensity than digital “0” (Figure 2d). Through synchronization, each micro-coil encodes digital data frame-by-frame so that neural data from each device can be localized and extracted from dynamic MR images.
Proof-of-concept:
i) The data uplink method was demonstrated with a 3T GE MR750W system (Waukesha, WI). A 2-mm double-loop coil was placed in a container with 3.3% NiCl₂(aq). The coil was controlled by a Teensy3.5 microcontroller (Figure 3a). Controlled currents (100uA-800uA) were passed through the coil, and 2D-coronal images through the phantom at varying TE (12ms-100ms) were taken.
ii) To demonstrate feasibility of device miniaturization, several parts of the envisioned system were designed. A 0.7mm thick PZT-4 sheet was diced into 1.54x1.54mm² cuboids, bonded to a flexible PCB, and encapsulated by 3.5um thick Parylene C (Figure 4a). The device was characterized by a Network Analyzer in a DI water tank while receiving ultrasonic power from a 1.18MHz transducer (Figure 4b). An IC containing a micro-coil for data uplink, a rectifier and demodulation circuit for ultrasonic energy harvesting and data downlink was designed in TSMC 28nm technology (Figure 5), and is currently being fabricated.

Results and Discussion

Simulation shows that with a 650um 10-turn coil, 100uA current and 60ms TE, 15% signal modulation can be achieved with SE-EPI (Figure 2c). Contrast can be increased by optimizing TE, or increasing coil current at the cost of power consumption. Experimental results show that in GRE-EPI, generated contrast competes with natural T2* decay (Figure 2d); this is improved by using bipolar coil current in SE-EPI. Figure 3d shows data “110101” encoded in images in an SE-EPI scan. With increased TE and coil current, better signal contrast was achieved, matching simulation results (Figure 3b,c).The PZT can harvest 95uW at the FDA limit for diagnostic ultrasound (720mW/cm²), sufficient for powering the IC. To program the mote while powering, we demonstrate amplitude modulation at 300Hz with 50% duty cycle (Figure 4d).

Conclusion

We propose the concept of a wireless deep brain neural implant that harvests power from focused ultrasound and uses a micro-coil to achieve simultaneous data uplink at multiple sites via dynamic MRI. We conducted proof-of-concept experiments to demonstrate data encoding feasibility of a micro-coil, and energy harvesting capability of a piezo.

Acknowledgements

The authors would like to thank the sponsors of the Berkeley Wireless Research Center and TSMC university shuttle program for chip fabrication. We thank the following funding source: National Institutes of Health (NIH) grant R01MH127104.