Computational Microscopy for Implementation of SLAP2 Functional Imaging

Computational Microscopy for Implementation of SLAP2 Functional Imaging

The properties of neuronal information processing of individual brain neurons, involving dendritic integration of synaptic inputs and encoding of resulting transformations in action potential output is poorly understood. The slow speeds of conventional in vivo imaging technologies, like galvo-mirror based laser scanning two-photon microscopy, for capturing fluorescent biosensors of neural activity are incapable of fully tracking information throughout complex 3D neuronal dendritic arbor structures. We have increased 3D imaging rates with our development of an Acousto-optic deflector (AOD)-based random access two-photon microscope (Sakaki, 2020), but full-neuron acquisition rates are limited to ~20 - 30 Hz. A major advance in this field has been achieved by Dr. Kaspar Podgorski’s recent development of the Scanned Lined Angular Projection (SLAP2) microscope capable of kHz sampling of 2D planes and >150 Hz 3D acquisition rates. Full realization of SLAP2, however relies on computational microscopy for rapid identification of neuronal elements within 3D brain volumes, and selection of regions of interests (ROIs) for conducting ultra-fast activity sampling. As the morphology of neural structures change over time due to growth and plasticity, repeated rounds of neural segmentation and ROI selection must be conducted throughout each experiment. Here, we describe the function of SLAP2, and highlight the numerous technological advances in its subcomponents. Further, we present our machine learning approaches for computational microscopy for fast neuronal segmentation and ROI selection.