Implantable computers

Chris Diorio

Neurons and neuronal networks decide, remember, modulate, and control an animal’s every sensation, thought, movement, and act. The intimate details of this network, including the dynamical properties of individual and populations of neurons, give a nervous system the power to control a wide array of behavioral functions; the unknown character of much of this detail motivates many workers in modern neurobiology. Neurobiologists examine the activities of brain cells tied to 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. They want to know more about neuronal dynamics and networks; about synaptic interactions between neurons; about how neuronal signaling and behavior and control and environmental stimuli are inextricably linked.

To make significant progress, neurobiologists need methods for recording the activity of single neurons or assemblies of neurons, for long timescales, at high fidelity, in animals that are free to interact with their sensory world, and that are free to express normal behavioral responses. Contemporary tools for studying neuronal signaling and information processing include imaging at the cellular and anatomical levels (e.g. single-photon microscopy, voltage-sensitive dyes, PET scans, etc.), direct electrical measurement using micropipettes or micromachined extracellular probes, and a host of other techniques. But the one tool that can truly help unravel the neural substrates of behavior—the digital computer—is, at present, utterly dissociated from nerve tissue. Neurobiologists use computers for simulation and for data collection, but not to probe the neurobiology directly—that is, they have not linked the electronic signaling of digital computers directly with the electronic signaling of nerve tissue.

Our research focuses on direct electrical measurement (and stimulation) of nerve tissue, primarily because voltage signaling is the lingua franca of both nerve tissue and computers—both use time-varying voltage waveforms to convey and to process information. Neurobiologists employ two classes of electrodes to record and stimulate electrical signals in tissue: (1) intracellular micropipettes for single-site intracellular and patch-clamp interrogation of neurons (one site per pipette), and (2) extracellular wires or micromachined probes, for interrogating multisite patterns of extracellular neural signaling or electrical activity in muscles. For recording in neurons, intracellular probes afford roughly a thousand-fold larger signal-to-noise ratio than do extracellular probes. However, multi-site recording is key to gaining access to patterns of neuronal activity. Currently, high-fidelity recording of neuronal signals is practical only on small numbers of neurons, in constrained laboratory experiments using immobilized animals, primarily because there are no probes for multi-site intracellular recording. For recording in muscles, extracellular probes allows adequate signal fidelity; the goal is to record or stimulate during normal behavior, so as to investigate the neuromechanical systems that underlie the behavior.

Tom Daniel , Denice Denton , Karl Böhringer , Dennis Willows , and I have initiated a multi-disciplinary research program to take the first steps toward integrating computer electronics with neurobiology. More specifically, we are (1) implanting a standalone microcomputer into the brain of a marine mollusc, to allow multi-site intracellular recording and stimulation in a live, freely behaving animal, and (2) embedding a standalone microcomputer into the sensory and muscle/control pathways of a giant moth, to probe the dynamical control of flight in a live, freely behaving animal. The compelling scientific reason for this research is to correlate neuronal signaling and control with environmental stimuli and behavior, to better understand the neural substrates of behavior.

Our research comprises four main thrusts: (1) developing a standalone implantable microcomputer that records from and stimulates multiple neurons or muscle fibers in an intact animal, using intracellular probes, extracellular probes, or wire electrodes; (2) developing micro-electromechanical (MEMS) probe arrays for recording intracellularly from multiple neurons simultaneously ; (3) developing neurophysiological preparations and techniques for implanting microchips and wire electrodes or MEMS probes into or onto animals without damaging probes or tissue, including dealing with issues of biocompatability, and (4) developing analytical models of the biology, and experiments that software-test these models in behaving animals.

Our research focuses on Manduca sexta and Tritonia diomedea, for several reasons. Both presently serve as model systems for probing the neural bases of behavior. Both afford accessible recording sites, with high signal fidelity. Manduca's flight circuits are relatively well understood in the context of constrained laboratory environments. Tritonia has extraordinarily large brain nerve cells, identifiable sensory and motor functions associated with these brain cells, and robust response to surgical insult. And, appearances to the contrary, the microcomputers that we are developing will be compatible with both animals.

A. Manduca Sexta                                        B. Tritonia Diomedea

Manduca Sexta is typically 4cm in length, with a 12cm wingspan; at 2.5g, it is among the largest of insect flyers. It can easily carry a test-electronics payload. Tritonia Diomedea is typically 20cm in length, and has a readily accessible brain with large and well-characterized neurons.

back to Chris Diorio's home page