Cyborg Insects and Other Things
February 13, 2014, Berkeley EECS Annual Research Symposium, UC-Berkeley, CA—Michel Maharbiz discussed research into interfaces between synthetic and multicellular objects and neural dust. These cyborgs—bugs with computer interfaces—allow researchers to stimulate the brains and control motion.
During some of their testing, they found that their bugs had other light sensors in addition to their compound eyes. These sensors detected UV and created an artificial horizon. Stimulating these cells affected body roll, which caused untethered insects to change direction in attempts to keep the horizon level. In the development for their cyborgs, they had to develop the technology for the brain stimulators that were less than 150 milligrams including the power source, resulting in a concept they called neural dust.
This development led to other work in recording spatial-temporal activity maps of the human brain. This is orders of magnitude more difficult than insects. To get reasonable fidelity, the planned to map a million points. The human brain is composed of billions of neurons that are organized in many levels of hierarchy. The neurons work on a electro-chemical input-output function and all activity has to be recorded outside the cell.
The electrical potentials are very small and mixed in with millions of other neurons firing at the same time. In general, it is not realistic to work with single neurons to catch spikes, because of the inter-neural interactions. Recording takes a long time and the electrodes have to be bio-compatible to minimize infections. Previous work used arrays of nano-needles, that caused insertion damage and needed to be wired to collect data.
The needle arrays don’t scale to large numbers, and the array is rigid and fairly invasive. The largest, wired needle arrays are about 250 x 450 microns and require about 2.5 microwatts per channel. A different querying technology is to use ultrasound, which is much more efficient for power and data transfers than electrodes. A piezo-electric sensor is placed in the head and changes backscatter and resonant frequency when an adjacent neuron fires.
The ultrasound can couple energy with about 7 percent efficiency due to the low velocity of the sound compared to electricity or RF. The sensor array uses maximum spacing for the electrodes to increase the differential voltage response. The biggest challenge is the thermal noise and scaling the mote to a 50 micron lower limit. The single FET sensor detects a neuron spike, which changes the FET impedance backscatter.
The power coupling need tight impedance matching to resonate and return a chirp big enough to detect. Current designs are able to detect a change of one part per million. Initial tests in animals used 100, 50-nm motes in a cube with a polymer tail. The tail has a conductor on it of about 1-5 microns that help separate the electrodes from the rest of the structure.
The 100 micron untethered modes operate on ultra-low power and can scale up to many sensors by frequency binning. Increasing the number of elements and phasing power to different nodes enables beam steering to interrogate sensors in an area.
