CaliAli Parameters Overview

This table lists all CaliAli parameters, their default values, a brief description, and guidance on how to choose them.

📌 Parameter Table

Parameter Name	Default Value	Description	How to Choose
`gSig`	`[] (auto)`	Neuron filter size in pixels	Defaults to `5 / spatial_ds`. Override when your data are anisotropic or when neuron diameters differ significantly from 5 px. Use NeuronSize_app for fine tuning.
`sf`	`10`	Frame rate (fps)	Set to match the acquisition frame rate.
`input_files`	`[]`	Paths to input video files	Leave empty to manually select files.
`output_files`	`[]`	Paths to output video files	Leave empty for default naming (recommended).
`batch_sz`	`'auto'` (demo) / `0`	Automatic chunking for large sessions (frames per chunk, 0 = disable)	Auto-estimates chunk size from available RAM; set a numeric value or `0` to override.

Parameter Name	Default Value	Description	How to Choose
`BVsize`	`[]`	Size of blood vessels in pixels [min, max]	Leave empty to automatically calculate based on gSig or use BV_app (recommended).
`spatial_ds`	`1`	Spatial downsampling factor	Increase for faster processing, decrease for higher resolution.
`temporal_ds`	`1`	Temporal downsampling factor	Increase only if memory constraints prevent full processing.
`file_extension`	`'avi'`	File extension for processed videos organized in folders	Used when sessions are split into multiple files.
`force_non_negative`	`1`	Enforce non-negative pixels during preprocessing	When enabled, values are lifted and clipped within `CaliAli_remove_background` after noise scaling.
`force_non_negative_tolerance`	`13`	Non-negative tolerance threshold	Gap added before clipping; increase only if you observe residual bias in dark regions.

Parameter Name	Default Value	Description	How to Choose
`neuron_enhance`	`true`	Use MIN1PIE background subtraction	Keep enabled unless signal loss is observed during preprocessing.
`noise_scale`	`true`	Enable noise scaling per pixel	Keep enabled unless signal loss is observed during preprocessing.
`detrend`	`1`	Detrending window (seconds), 0 = no detrending	Set to the duration of calcium transients in seconds.
`median_filtering`	`[]`	Median filter window `[rows, cols]` applied per frame	Enable when noise scaling creates hot pixels near vignetted borders; `[3, 3]` is a good starting point.

Parameter Name	Default Value	Description	How to Choose
`reference_projection_rigid`	`'BV'`	Reference projection for rigid correction	Choose `neuron` if blood vessels are not suitable.
`do_non_rigid`	`false`	Perform non-rigid motion correction	Enable only after confirming rigid correction was insufficient.
`non_rigid_pyramid`	`{'BV','neuron','neuron'}`	Multi-level registration pyramid for non-rigid correction	Use default unless BV is unavailable.
`non_rigid_batch_size`	`[20,60]`	Batch size range for non-rigid correction, CaliAli will optimize within this range.	Set as `[2 x sf, 6 x sf]`.

Parameter Name	Default Value	Description	How to Choose
`do_alignment_translation`	`true`	Perform inter-session translation	Always true unless sessions were pre-aligned. If do_alignment_non_rigid is also false, videos will be concatenated without registration.
`do_alignment_non_rigid`	`true`	Perform inter-session alignment	Always true unless sessions were pre-aligned or non-rigid alignment is not necessary. If do_alignment_translation is also false, videos will be concatenated without registration.
`projections`	`'BV+neuron'`	Projection method used for alignment	By default use both blood vessels and neurons.
`final_neurons`	`false`	Use an additional alignment iteration based on neurons	Enable if session registration was inaccurate. Often not necessary
`Force_BV`	`0`	Force BV alignment even if stability score is low	Sometimes BV stability score may be low as results of a debris in the FOV. Setting this parameter to true would ensure that BV are used for registration

Parameter Name	Default Value	Description	How to Choose
`memory_size_to_use`	`total_system_memory_GB` (auto)	Total available memory for computation	Override when you want MATLAB to use less than the detected RAM.
`memory_size_per_patch`	`total_system_memory_GB` (auto)	Memory allocated per patch	Defaults to the detected RAM so patching adapts to your hardware; reduce if you need smaller tiles.
`patch_dims`	`[64, 64]`	Dimensions of patches	Larger patches improve accuracy but increase computation time and memory consumption.
`w_overlap`	`32`	Patch overlap width	Increase if you detect edge artifacts.

Parameter Name	Default Value	Description	How to Choose
`min_corr`	`0.1`	Minimum correlation for neuron seeding	Usually controlled via `CaliAli_set_initialization_parameters(CaliAli_options)`.
`min_pnr`	`6`	Minimum peak-to-noise ratio for seeding	Same as `min_corr`.
`min_pixel`	`[]`	Minimum pixel area for neurons	Automatically calculated based on gSig.

Parameter Name	Default Value	Description	How to Choose
`with_dendrites`	`true`	Do not assume that signals are circular during extraction	Keep enabled for better CNMF convergence and accurate dendrite detection.
`spatial_constraints`	`{'connected': False, 'circular': False}`	Constraints for spatial filtering	Always disable for better CNMF convergence and dendrite identification.
`spatial_algorithm`	`'hals_thresh'`	Algorithm for spatial extraction	Use default unless alternative extraction methods are needed.

Parameter Name	Default Value	Description	How to Choose
`deconv_flag`	`true`	Enable deconvolution	Use default if unsure. Enable for better temporal resolution.
`deconv_options`	`{'type': 'ar1', 'method': 'foopsi', 'smin': -5, 'optimize_pars': True, 'max_tau': 100}`	Deconvolution settings	Use default parameters to minimize false-positives and computational stability

Parameter Name	Default Value	Description	How to Choose
`merge_thr_spatial`	`[0.8, 0.4, -inf]`	Merge components with highly correlated spatial shapes (`corr=0.8`), moderate temporal correlations of calcium activities (`corr=0.4`), and disregard spikes correlations (`corr=-inf`).	Increase the spatial correlation threshold when you need stricter merging so only very similar components combine.