moco: Fast Motion Correction for Calcium Imaging

Abstract

Motion correction is the first step in a pipeline of algorithms to analyze calcium imaging videos and extract biologically relevant information, for example the network structure of the neurons therein. Fast motion correction is especially critical for closed-loop activity triggered stimulation experiments, where accurate detection and targeting of specific cells in necessary. We introduce a novel motion-correction algorithm which uses a Fourier-transform approach, and a combination of judicious downsampling and the accelerated computation of many L 2 norms using dynamic programming and two-dimensional, fft-accelerated convolutions, to enhance its efficiency. Its accuracy is comparable to that of established community-used algorithms, and it is more stable to large translational motions. It is programmed in Java and is compatible with ImageJ.

Keywords: Machine Vision Algorithms 150.1135; calcium imaging; dynamic programming; fourier transform; mesoscale neuroscience; motion correction.

Figure 1

Image registration with moco. 1.a. and 2.a. are first two images of a long, badly corrupted video submitted to moco. 1.b. and 2.b. are the two corrected images. Note that 1.a. and 1.b. are identical, since 1.a. is used as the template image. However, 2.b. is registered by moco, moved to the right to overlap 1.b., it matches it almost perfectly except for the non-overlapping black rim. Images are 317.44 × 317.44 mm.

Figure 2

Image registration with two comparable algorithms. 1.a. is the mean of all frames in a badly corrupted video. 2.a. is the of corrected video using our implementation of the (Guizar-Sicairos et al., 2008) approach. 1.b. is the mean of the corrected video using moco. 2.b. is the mean of the corrected video using TurboReg (accurate mode), Thevanaz et al. (1998) 1.c. is the mean of the corrected video using TurboReg (accurate mode). 2.c. is the mean of the corrected video using Image Stabilizer. Note that moco and TurboReg have superior performance, as noted by the sharper and brighter appearance of the cell bodies. Images are 317.44 × 317.44 mm.

Figure 3

Analysis of spurious translations by moco and a similar algorithm. Left and right images i right show differences in displacement in the first and second dimensions as a function of time in moco and (Guizar-Sicairos et al., 2008). The video on which they were applied was a real calcium image with added horizontal and vertical spurious translations, to make the task more difficult. moco and (Guizar-Sicairos et al., 2008) generate different translations, and the differences in the translations found are plotted. The left plot shows the y-translations that moco makes minus the y-translations that (Guizar-Sicairos et al., 2008) makes. This difference is typically zero, but there are notable exceptions. The right plot does the same thing for x-translations. Note how moco detects many more translations.