Click here to login if you're an NAE Member
Recover Your Account Information
Author: Chris Diorio
Linking Artificial Information-Processing Machines and Living Information-Processing Machines
Neurobiologists want to understand how neurons control animal behavior.
A key goal of neurobiology is understanding how neurons control behavior. Neurobiologists probe and examine the activities of brain cells involved in sensory inputs, integrative processes, and motor outputs to understand the neural bases of behavior. They also probe the components of neuronal control circuitry to understand the plasticity and dynamics of control. But to make significant progress, neurobiologists need methods of recording the activity of single neurons or assemblies of neurons for long time scales at high fidelity in animals that are free to interact with their sensory world and free to express normal behavioral responses.
Although both silicon electronics and nerve tissue process and communicate information using primarily electrical (voltage and current) signals, there has been no large-scale integration of silicon electronics with nerve tissue either for investigatory or medical purposes. Although the electronics for recording from and stimulating neurons already exists as bench-top equipment, it hasn’t been miniaturized in the form of a stand-alone, implantable computer. My goal is to build miniature computers that we can implant into or onto animals without damaging the computer or the animal that can interface directly with nerve tissue.
One step toward achieving this broad objective is understanding how animals control movement, a fundamental problem we can readily study because we can probe the structures, muscles, and nerves that enable movement. However, understanding neuronal control itself has remained beyond our grasp, partly because of our inability to study in sufficient detail the dynamics of this control in intact, freely behaving animals. Consequently, our first effort in implantable electronics is to study the mechanisms of sensorimotor integration and the control of locomotion in freely behaving animals.
This interdisciplinary project encompasses biologists, engineers, chemists, and others. Choosing the animal was easy; my colleague, Tom Daniel, already had a fantastic model system: Manduca sexta, or hawkmoth. Manduca is one of the largest flying insects, and its flight circuits are relatively well understood in the context of constrained laboratory environments. Tom and others have studied this animal extensively, including its flight dynamics, neuromuscular control, and visual and mechanosensory signaling. In addition, Manduca has accessible recording sites with high signal fidelity.
We had several options for computer-based studies: tethering the animal to a host computer; mounting simple electronics on the animal and communicating the recorded information via radio frequency to a host computer; or using a stand-alone microcomputer on the animal itself. We chose the third option, for several reasons. First, the technology is totally autonomous. Second, it can be used on other animals - for instance, we are using it underwater to study the nudibranch Tritonia. In addition, as electronics advances, the amount of computational circuitry on a single chip will increase amazingly, and implantable computers will become commonplace. Single-chip computers today have sufficient power for biologists to perform complex experiments on an animal in real time.
Our immediate goal is to choose a biology experiment, write the experiment in software, download the software to the implantable computer before the experiment, run the experiment on the untethered animal, and then upload the data to a desktop computer when the experiment is over. For the actual interfacing, the input side needs amplification and an onboard central processing unit - the little microcomputer (actually a microcontroller) - to run the experiment, plus onboard memory and a battery. The output side needs an interface to a desktop computer.
Eventually, we will increase our experimental capabilities. We’ll add an accelerometer to record flight forces on the animal as it flies. We’ll add onboard optical sensors so we can correlate the experiment with the high-speed video we are taking externally. And we’ll need outputs for direct muscle stimulation to do stimulus-response experiments.
The constraints on an implantable computer are size and power consumption. The electronics must operate at incredibly low current because the power comes from onboard batteries, and a moth cannot carry much weight. Our first prototype was a printed circuit board - a test bench for validating designs before we committed them to silicon. The board used off-the-shelf components, which are incredibly capable. We used a small programmable system-on-a-chip, basically a little microcomputer with some amplifiers and a bit of memory. We added more memory and amplifiers. Just for calibration, this tiny microcontroller on a tiny printed-circuit board is an 8-bit processor running at 24 megahertz, roughly the same power level as the old 8088s in the first IBM PCs. But the main challenge was and remains the battery. Our goal is power consumption of less than 15 milliwatts. Right now, we are using a zinc silver-oxide battery because it works in any environment, but it has some disadvantages - low current density, short lifetime, and heavy packaging. Eventually, we would like to have thin-film, rechargeable batteries, which may be available commercially soon, that could handle an extended experiment that includes recording and stimulating.
Our first proof-of-concept implant was not very elegant, but we learned a lot from it. The little board we used was about 1 by 3 centimeters and weighed about 1.5 grams - too heavy for the moth to fly with. Still, we discovered that by having the amplifiers and the batteries right on the moth we got incredible signal fidelity at very low power consumption because the length of the wires from the amplifiers into the moth was so short. Although the moth couldn’t lift this prototype, that wasn’t the purpose. The purpose was to integrate all of the electronics on a little board, put it right next to a moth, and run actual experiments collecting information, compressing and storing it, and later downloading it to a PC.
Our next goal is miniaturization, both in size and in power consumption. By the time you are reading this article, we will have a board about 1 centimeter in diameter, a little hybrid ceramic substrate - no packaging, no soldering, no plastic - just chips on a board, weighing roughly 0.75 gram including batteries. The animal will be able to fly with this board glued to its thorax. We will conduct implant experiments and record during free flight. Our future goal is to put all of the capabilities on a single chip and eliminate the ceramic board entirely. Using current technology, that chip would be about 6 millimeters by 6 millimeters and would weigh about 0.4 grams with a battery.
In addition to the implantable electronics, my colleague Karl Bohringer is developing intracellular silicon probe arrays with micromachined submicron tips to penetrate and record from inside nerve cells. Unlike glass pipettes that require bench-top micromanipulators for placement, his arrays will be stand-alone devices that attach to the neural ganglion using microfabricated hooks or clamps built on the array itself. When we combine our implantable computers with his microprobes, we will be truly able to study and explore integrative processes within animal brains. Dr. Bohringer’s work is incredibly challenging, but, if it is successful, it will change the way we do biology.
As we begin to explore animals in their natural environments using stand-alone computers, many technologies will come together: low-power microelectronics; new battery technologies; MEMS probes; new surgical procedures; and software algorithms. Initially, the goal is to record from the animals. Eventually, we hope to close the loop - to stimulate the animal and measure its response. In short, we want to build a smart electronic system that interacts in a smart way with an animal brain. The possibilities that derive from linking artificial information-processing machines (computers) to living information-processing machines (animal brains) defy prediction.
This research was funded by the Office of Naval Research and the David and Lucile Packard Foundation.