Appendix 1—table 3. List of algorithm implementations in different pipelines.

For a lot of steps, different algorithm implementation can be chosen by the user based on features of the data. In such cases, we only list the default and most commonly used algorithms here.

Step	Minian implementation	CaImAn implementation	MIN1PIPE implementation	Critical parameters
Denoising	Median filter	None	Anisotropic filter	Spatial window size of the filter
Background removal	Morphological top-hat transform	None	Morphological top-hat transform	Spatial window size of the top-hat transform
Motion correction	FFT-based translational motion correction	Nonrigid patch-wise translational motion correction (NoRMCorre)	Mix of translational motion correction and Demons diffeomorphic motion correction	Different
Initialization	Seed-based with peak-noise ratio and KS-test refinement	Pixel-wise correlation and peak-noise ratio thresholding	Seed-based with GMM, peak-noise ratio and KS-test refinement	Threshold for correlation and peak-noise ratio
Spatial and temporal updates	CNMF with CVXPY as deconvolution backend	CNMF-E with Oasis as deconvolution backend	CNMF with CVX MATLAB package as deconvolution backend	Noise cutoff frequency Expected size of neurons Sparse penalty

GMM: Gaussian mixture model; KS: Kolmogorov–Smirnov; CNMF: constrained non-negative matrix factorization.