DustNet: A Wireless, Battery-Free, Sub-mm Scale Neural Recording Implant Network
The most commonly used method for in situ neural interfaces involves implantation of micro-electrodes on deeply-seated nerves that are connected to
an external device, using transdermal wires, which performs recording and decoding of neural data to infer intention. The tethered nature of this method
results in a relatively short implant lifetime (a few weeks) due to infection, displacement of the implanted electrodes caused by externally applied force
on the wires and subsequent signal quality degradation, all of which ultimately limit the application of such implants to basic neuroscience research. To
increase the longevity of peripheral nerve implants and extend their application beyond neuroscience research, we propose a network of battery-free
sub-mm scale implants that wirelessly communicate with an external interrogating device, as shown in Figure 1, eliminating wires, and substantially
reducing the likelihood of infection and implant displacement.
Our lab has previously developed miniaturized wireless, free-floating neural implants for neural recording and stimulation. These devices rely on
piezoelectric transducers to harvest power from an external ultrasound transducer and modulate the impedance of the piezo to communicate data via the
backscattered acoustic wave. These two implants, however, operate as standalone devices. In this work, we propose to use an ensemble of these two
implant types to enable minimally invasive closed-loop neuromodulation for peripheral nerve targets. The main requirement for such a network is a
multiple-access wireless protocol to allow bi-directional communications with multiple recording and stimulating implants. Increasing the data rate in
ultrasonic implants to support multiple implants is fundamentally difficult because ultrasound communication occurs in a pulse-echo manner with short
pulses that makes it difficult to modulate large amounts of information - in other words, the channel capacity is quite low.
To realize a neuromodulation system free of these constraints, To implement sensor nodes for a large-scale neural recording network, we have developed
a second-generation neural recording implant that exploits our understanding of the nonlinear properties of ultrasound backscatter to increase channel
bandwidth and enable high-speed, multi-node data communication. These “Neural Dust” sensor nodes use a time-division multiple-access (TDMA) protocol
to enable bidirectional communication between an external transducer and multiple implants. In contrast to previous implants that rely on OOK modulation
for data communication, the Neural Dust nodes use an M-level ASK backscatter modulation scheme to increase the channel bandwidth by a factor of
log2(M). These sensor nodes incorporate active rectifiers to harvest power from the piezoelectric transducer, a low-noise neural recording front-end, active
power management unit, and modifies the piezo termination impedance to backscatter data.
The prototype implant IC was fabricated in a TSMC 28 nm HPM CMOS process, which consumes 5 μA of current during operation and occupies an area of
1×0.4 mm2. The second-generation DustNet chip was taped-out in August 2023. Now, we are focusing on characterization of the prototype implant IC
using a bring-up board. It is essentially an arbitrary waveform generator specifically tailored to generate the piezo signal that the chip would receive from
the implant piezo when powered acoustically. After validating the functionality of the prototype implant, these devices will be assembled and tested with
collaborators in mammalian peripheral nerves to demonstrate recording fidelity from nerve bundles and stimulation capabilities. Future research will be
focused on combining this work with stimulation implants to create a true closed-loop neuromodulation network.
Expected Graduation Date:
May, 2028
