Hardware Efficient Image Reconstruction for Two-Photon Microscopy

Two-photon excitation microscopy is a fluorescence imaging technique that can image deeply through living tissue in a minimally invasive fashion, establishing itself as a mainstay for in-vivo neuroscientific investigation. We aim to miniaturize an entire two-photon microscope to enable experimentation in freely moving organisms, requiring miniaturization of the electronics/optics as well as hardware specific software to control it. Processing the raw data coming from our system is a job poorly suited for a CPU; the associated computations use up very few hardware resources, so we investigate the use of highly parallelized architectures for the potential speedups. We use NVIDIA’s CUDA library to design our algorithms, and we process recorded data sets to demonstrate the reduced processing time and corresponding raw throughput rates.