How neural activity can be translated into interpretable visual content
Summary
Brain reading is an emerging direction in our lab: we aim to translate patterns of neural activity into testable descriptions or reconstructions of visual experience while determining exactly what a successful decoder has learned. Our large-scale survey of the mouse visual system maps the representations available across areas, our work on recording-modality differences clarifies which neural populations a decoder can observe, and our digital-twin perspective describes how predictive models can be interrogated for selectivity and invariance. We now ask which aspects of visual content can be decoded reliably, whether decoders transfer across animals, sessions, stimuli, and brain states, and how much of a reconstruction comes from neural evidence rather than the model’s prior knowledge.
From the lab
Foundational publications from our group
Brain reading is an emerging lab program. These publications provide its experimental and modeling foundations; they are not presented as completed brain-to-image reconstruction studies from our group.
Mapping neuronal selectivity and invariance in the mammalian visual cortex using digital twins
Defines how predictive digital twins can be interrogated to recover meaningful dimensions of neural response.
Survey of spiking in the mouse visual system reveals functional hierarchy
Provides a large-scale map of what can be read from spiking activity across the mouse visual system.
Reconciling functional differences in populations of neurons recorded with two-photon imaging and electrophysiology
Quantifies how recording modality changes the neural population available to an encoder or decoder.
Systematic Integration of Structural and Functional Data into Multi-Scale Models of Mouse Primary Visual Cortex
Builds a biologically constrained model that connects circuit structure to visual responses.