Neuroengineering & Systems Neuroscience
|
Research |
|
The RoLi lab is jointly led by Drew Robson and Jennifer Li. We seek to discover computational principles and mechanistic implementations of cognition in both biological and artificial systems.
We work with the larval zebrafish, a unique model organism with a compact and transparent brain in which nearly every neuron can be simultaneously recorded. Our neuroengineering team develops novel imaging systems to record and manipulate neural activity in freely swimming larval zebrafish, and our systems neuroscience team uses these tools to gain a deeper understanding of the neural mechanisms that organize internal brain states and abstract cognitive representations. Our lab is funded by the Max Planck Institute for Biological Cybernetics and the European Research Council (ERC Synergy Grant 2025-2031) 
|
|
Neuroengineering Conventional microscopes are stationary passive imaging systems, where the animal must be restrained. This significantly restricts the range of natural behaviors the animal can perform. Our lab creates imaging systems that actively track the animal, using robotics and real-time computation to keep the animal's body and brain in focus. These are a new generation of self-driving or autonomous microscopes, which make decisions in real time in response to the behavior of a living animal.
The tracking microscope
Our first-generation tracking microscope was the founding technology of the lab. By combining whole-brain imaging with a robotic motion-cancellation system, it became the first instrument capable of recording activity from across the brain, at the level of individual neurons, in a freely swimming zebrafish (Kim et al., Nature Methods, 2017). This single tool reshaped what is possible in zebrafish systems neuroscience, and led directly to several of our biggest discoveries, including brain states that govern foraging and the discovery of place cells in fish. 
Kim DH, Kim J, Marques JC, Grama A, Hildebrand DGC, Gu W, Li JM* & Robson DN*. Pan-neuronal calcium imaging with cellular resolution in freely swimming zebrafish. Nature Methods 14(11), 1107–1114 (2017).
DOI: 10.1038/nmeth.4429 | Read online Foveated imaging (DASHER)
This system dynamically adjusts the region(s) of interest on a camera for recording or analysis, enhancing the spatial and temporal resolution around each animal in motion while preserving a wide field of view. Using DASHER, we can resolve fine features such as eye movements across many animals at once. It revealed that fish sleep divides into conserved substates with distinct eye-movement kinematics (Choudhary, Heller et al., Nature Communications, 2026; New York Times: Fish Sleep a Lot Like Us. They Even Nap.) Upcoming ...
Continuous circadian imaging: A gentler, oblique light-sheet design cuts excitation power by more than an order of magnitude, enabling the first uninterrupted whole-brain recording of a freely behaving vertebrate across day and night. Reading and writing neural activity: Watching the brain is only half the goal. To prove that a circuit causes a behavior, we need to perturb it directly and see what changes. We are developing dual read/write microscopes that combine imaging with precise optical manipulation of chosen neurons (for example, specific place cells) while the animal behaves. Closing this loop, from observation to intervention to response, lets us turn correlations into mechanisms. Towards autonomous life-time imaging: We are creating an autonomous imaging platform with onboard life support, environmental control, and adaptive decision-making that can record from the same animal for weeks, without human intervention. This makes a genuinely new kind of experiment possible: following a single brain continuously as its circuits and cognitive abilities emerge and mature. |
Systems Neuroscience Using our unique technologies, we seek to discover how the brain can efficiently and flexibly create internal representations of the external world. With only ~100,000 neurons total at larval stages, the zebrafish brain is a remarkable example of how a small neural network can efficiently store a vast amount of information about the world, including spatial cognitive maps, environmental context, and social interactions.
We aim to build a mechanistic model of cognition in this tractable animal. Spatial cognition: How a brain builds a map of the world
Knowing where you are and how to get from A to B is one of the most fundamental forms of cognition. The brain does this by maintaining an internal cognitive map, and a key ingredient is the place cell, a neuron that fires only when the animal is in a particular location. Using the tracking microscope, we uncovered place cells in the zebrafish, the first systematic discovery of such cells in a non-amniote vertebrate (Yang et al., Nature, 2024). 
Illustration by Gon Yishay on the discovery of place cells in fish
Yang C, Mammen L, Kim B, Li M, Robson DN & Li JM. A population code for spatial representation in the zebrafish telencephalon. Nature (2024).
DOI: 10.1038/s41586-024-07867-2 | Read online These maps are thought to arise from recurrent networks. Compared with simpler feedforward networks, the operations of biological recurrent network are still poorly understood. Because the zebrafish brain is small enough to study in its entirety, we can do something almost impossible in larger brains: combine, in one animal, the neural activity, the wiring diagram, and the molecular identity of the cells that make up a complete cognitive network. This provides a rare opportunity to build a brain-wide mechanistic models of spatial cognition.
Neuromodulation: the brain's changing internal states
The same brain behaves differently depending on its internal state, whether it is hungry or sated, awake or asleep, exploring broadly or focused on a task at hand. These shifts are largely driven by neuromodulators, chemical signals that reconfigure how circuits operate. We study how the dynamics of these internal states organize decision-making, attention, and memory. Two recent examples from our lab:
Illustration by Vikash Choundhary: eye movements during a sleep substate (quiescence with eye movement type 1, QEM-1)
A central theme of our ongoing work is to understand how neuromodulation helps to construct and reorganize cognitive representations, acting as a signal that shapes learning and memory as circuits change.
Development: how cognitive abilities are assembled
Cognition is not present fully formed at birth; it is assembled. Yet we rarely get to watch the assembly happen, because tracking the same neurons over the long stretches of development has been technically out of reach. We are creating an autonomous imaging platform with onboard life support, environmental control, and adaptive decision-making. By recording a single brain continuously as the animal matures, we aim to directly observe the emergence, stabilization, and reorganization of spatial and neuromodulatory networks. A finding that may be invisible in the finished adult brain (how a cognitive map is first built, and what role neuromodulation plays in building it) becomes visible during the open construction phase of the young zebrafish. |