Imaging three dimensional visually-evoked calcium dynamics in larval zebrafish optic tectum using light field microscopy

Networks of neurons give rise to behavior in complex animals. Determining the relationship between neural activity and sensory input or motor output requires recording network activity over time. Two properties contribute to the difficulty of recording from intact biological neural networks: first, the component neurons are spread across the three-dimensional volume of the brain, and second, all neurons in the network may be active simultaneously. Recently, optical techniques have enabled monitoring activity across many neurons simultaneously. However, the recording range has been limited to the field of view of imaging device, generally the 2D focal plane of a camera or scanning laser system. In order to capture activity information across the full 3D extent of the network, multiple exposures must be acquired at different focal distances, which breaks the synchrony of the activity measurements. Here, I describe a novel neural imaging technique, called light field microscopy (LFM), which is capable of simultaneously recording the activity of many neurons across a volume of tissue. By measuring both the spatial and the angular properties of incoming photons, LFM is able to capture a 3D volume with a single camera exposure. To demonstrate the efficacy of LFM for neural recordings, I loaded the optic tectum of the transparent larval zebrafish with fluorescent calcium-indicator dye and presented the fish with high-contrast 2D visual stimuli. Using this preparation, LFM produces a 4D dataset (three spatial dimensions plus the time dimension) representing the visually-evoked activity patterns across a 3D network of neurons. In this dissertation, I describe the steps required for producing 4D neural activity datasets, and compare the resulting data to that obtained using a classic two-photon scanning laser microscope. The wealth of data captured by the LFM increases the complexity of analysis, and thus motivated the development of a statistical method for automatically isolating interesting parts of the dataset. The method, called elastic source selection (ESS), can determine the relevance over time of each captured voxel to an external stimulus. This dissertation establishes the effectiveness of the LFM and ESS techniques, providing neuroscientists a new avenue into the brain.