Fig. 2: Performance of DNN-1 for SNR enhancement of in vivo head-fixed 2P fiberscopy images.
From: Deep-learning two-photon fiberscopy for video-rate brain imaging in freely-behaving mice
a Representative input image along with the magnified view (b) of the selected region. Images were acquired at a scanning speed of 3,360 spirals/sec and an imaging density of 512 spirals/frame. c Corresponding DNN-1 output image of (a) along with the magnified view (d). e Neuron identification by using a post-processing pipeline. f Normalized time-dependent fluorescence intensity (ΔF/F) of a selected neuron before and after DNN-1 enhancement. g Dynamic signal fidelity measured by normalized root-mean-square error (NRMSE) for each neuron. The differences of calcium signals between the DNN-1 output and the raw data are small (with an average discrepancy around 3%). Scale bars in (a, c): 20 µm.
