McCormick Magazine

Tackling the complex system of animal life

Researchers are developing a neuromechanical model of fish


Neelesh Patankar and Malcolm MacIverWhen the sun goes down, the black ghost knifefish trawls the muddy rivers of the Amazon Basin, using a long, ribbon-like fin on its underside to dart both forward and backward, and sending out electrical fields to find prey and navigate. The elusive fish are so mysterious that South American natives believed they housed the souls of departed friends and family.

But in a McCormick School lab, that mystery is rapidly disappearing. Researchers are working to create a neuromechanical model of the knifefish that could provide insight into how our brains take in information and use that information to move.

Malcolm MacIver, assistant professor of mechanical and biomedical engineering, has worked with the knifefish for some time, quantifying the signals acquired by the fish, creating computer models to reconstruct activity of its nervous system, and making robotic models. When he came to Northwestern in 2003, MacIver had already developed key portions of a computational model of the fish's sensory system.

Then two years ago, MacIver teamed up with Neelesh Patankar, associate professor of mechanical engineering, to develop a mechanical model that predicts how the knifefish moves in water. Patankar had already used such computational techniques to create an algorithm that simulates the motion of an object through a fluid. His model is so efficient that animators have even used it to program animated objects in water. Patankar's challenge with MacIver's knifefish was to extend that technique to a body that actively moved itself through water.

To do so, he devised a new algorithm to solve for both the swimming velocity of the fish (corrected so it's consistent with the movement of the fish's body and fins) and the velocity field of the surrounding fluid. That allowed him to calculate the force generated by the fish to swim.

Patankar led the effort in which he and MacIver worked with shared postdoctoral researcher Anup Shirgaonkar and PhD student Oscar Curet to create a mechanical model that, when given information on the fish's shape and how the fish moves its fins, can solve how that fish will swim and how the surrounding water will move. "Our algorithm can find a way to actually simulate how something goes from one location to another and what kind of forces the muscles would need to produce that movement," Patankar says.

When they applied the model to the knifefish, the team found that the knifefish's long fin pushes water out in rings of fluid that travel down its body — almost like it's swimming through a smoke ring. "Those rings carry momentum and generate force that pushes the animal in the opposite direction," Patankar says.

Now that the team has both a sensory model and a mechanical model of the fish, the biggest challenge lies ahead; they will combine the models to figure out what kind of message the brain sends to the muscle to make it move how it wants it to move. Because the fish brain is similar to the human brain, it can act as a simplified model and provide insights into how our own brains work.

KnifefishResearchers have begun developing computer models that integrate both areas, and MacIver and Patankar must figure out the mapping that happens in the brain between sensing and moving. "What happens in between is this really complex sensory motor transformation," MacIver says. "How that process works isn't really well understood because we're not approaching the system with any good hypotheses."

A combined model could give them a starting point to create such a hypothesis, he says. Then MacIver and his team will put electrodes on the fish's brain to test the hypothesis. "Through this collaboration we're going to get a first-of-its-kind high-fidelity neuromechanical model, which we can then use to explore very fundamental problems about the nervous system and how animals — including humans — take sensory data and generate motor output," MacIver says.

That information could be used to create better robots or ease movement disorders. It could also be used as a tool to understand the impact of mechanics on evolution, Patankar says, and could lead to better animation and technology in the gaming industry. "It would have all these applications that would make it a very good tool to have," he says.

—Emily Ayshford