PMID-28650477 Video rate volumetric Ca2+ imaging across cortex using seeded iterative demixing (SID) microscopy
- Tobias Nöbauer, Oliver Skocek, Alejandro J PernÃa-Andrade, Lukas Weilguny, Francisca MartÃnez Traub, Maxim I Molodtsov & Alipasha Vaziri
- Cell-scale imaging at video rates of hundreds of GCaMP6 labeled neurons with light-field imaging followed by computationally-efficient deconvolution and iterative demixing based on non-negative factorization in space and time.
-
-
- Utilized a hybrid light-field and 2p microscope, but didn't use the latter to inform the SID algorithm.
- Algorithm:
- Remove motion artifacts
- Time iteration:
- Compute the standard deviation versus time (subtract mean over time, measure standard deviance)
- Deconvolve standard deviation image using Richardson-Lucy algo, with non-negativity, sparsity constraints, and a simulated PSF.
- Yields hotspots of activity, putative neurons.
- These neuron lcoations are convolved with the PSF, thereby estimating its ballistic image on the LFM.
- This is converted to a binary mask of pixels which contribute information to the activity of a given neuron, a 'footprint'
- Form a matrix of these footprints, p * n, (p pixels, n neurons)
- Also get the corresponding image data , p * t, (t time)
- Solve: minimize over T subject to
- That is, find a non-negative matrix of temporal components which predicts data from masks .
- Space iteration:
- Start with the masks again, , find all sets of spatially overlapping components (e.g. where footprints overlap)
- Extract the corresponding data columns of T (from temporal step above) from to yield . Each column corresponds to temporal data corresponding to the spatial overlap sets. (additively?)
- Also get the data matrix that is image data in the overlapping regions in the same way.
- Minimize over
- Subject to
- That is, solve over the footprints to best predict the data from the corresponding temporal components .
- They also impose spatial constraints on this non-negative least squares problem (not explained).
- This process repeats.
- allegedly 1000x better than existing deconvolution / blind source segmentation algorithms, such as those used in CaImAn
|