Data from: A comprehensive suite for extracting neuron signals across multiple sessions in one-photon calcium imaging

Research facility: International Institute for Integrative Sleep Medicine

Published Mar 17, 2025 on Dryad. https://doi.org/10.5061/dryad.crjdfn3ck

Abstract

The dataset consists of calcium imaging data utilized for developing and validating CaliAli, a specialized suite for extracting neural signals from one-photon calcium imaging sessions under free-moving conditions across multiple sessions.

Key components of this dataset include:

Simulation codes used to generate synthetic data for this study.
Source data corresponding to each figure presented in the study.
The raw calcium imaging dataset employed in the research.

https://doi.org/10.5061/dryad.crjdfn3ck

The dataset consists of raw and processed calcium imaging data utilized for developing and validating CaliAli, a specialized suite for extracting neural signals from one-photon calcium imaging sessions under free-moving conditions across multiple sessions.

Description of Data and File Structure

The dataset is organized into the following categories:

video simulations: Scripts for generating video simulations.
Raw calcium imaging videos: Unprocessed imaging data.
Source data: Processed data supporting key analyses.
Extended source data: Large datasets not included in Source_data.xlsx.
Extracted neuronal signals: Processed neuronal activity data.
Codes: Codes to process video data.

Video simulations

Codes used to generate synthetic data for this study can be found in the Generate simulated data.zip, along with detailed instructions.

Raw Imaging Data

Raw calcium imaging videos for data produced in this study are included in .h5 format.

Folder Structure:

File Name (.zip)	Description
1h_CA1_recording	Fig S14. Tracking performance evaluation over short timescales.
DG_PVD_data	Fig 8, S15. Dentate gyrus recordings spanning 99 days.
Focal_plane_change	Fig S6, S9. Imaging sessions acquired at different focal planes.
Opto_data_30mins	Fig 7b-d, S13. Simultaneous imaging and optogenetic stimulation (30 min sessions).
Opto_data_4weeks	Fig 7e-g. Simultaneous imaging and optogenetic stimulation over 4 weeks.

Source Data

Quantitative data used for generating figures and conducting statistical analyses are stored in the Source_data.xlsx file. This file contains multiple sheets, each corresponding to a specific figure in the manuscript. Data are organized as follows:

Sheet Name	Description	Variables
Fig. 2e	Comparison of tracking algorithms based on correlation and crispness.	Algorithm (categorical), Correlation, Crispness
Fig. 3a	Effect of inter-session gap on tracking accuracy.	Inter-session gap (days), BV similarity score, Tracking accuracy (%)
Fig. 3d	BV similarity score and F1-score for neuron tracking.	BV similarity score, F1-score (%)
Fig. 4b	Spatial correlation of GCaMP vs tdTomato neuronal projections.	Correlation
Fig. 6d	# of neurons extracted by CaliAli vs. independent session processing.	Neuron count (%)
Fig. 6e	Place decoding error within sessions.	Decoding error (cm)
Fig. 6g	Percentage of neurons tracked across sessions.	Neuron count (%)
Fig. 6h	Place field correlation values.	Correlation
Fig. 6k	Decoding error across conditions (all neurons).	Decoding error (cm)
Fig. 6l	Decoding error across conditions (only tracked across all sessions).	Decoding error (cm)
Fig. 6m	Decoding error across conditions (downsampled to 100 neurons).	Decoding error (cm)
Fig. 7e	Opto-consistency in misaligned data.	Misalignment amplitude (µm), Opto-consistent neurons (%)
Fig. 7h	Opto-consistency across 4 weeks.	Neuron tracking method (categorical), Opto-consistent neurons (%)
Fig. 8c	Population vector distances of session pairs obtained across 99 days.	Population vector distance
Fig. S1b	Percentage of non-rigid misalignment in relation to the cumulative misalignment detected.	Misalignment (%)
Fig. S1c	Misalignment amplitude observed within a 4-day gap.	Misalignment amplitude (µm)
Fig. S1d	Misalignment amplitudes across time.	Misalignment amplitude (µm), Session gap (days)
Fig. S3e	Alignment scores performance using different projections.	Alignment score
Fig. S3f	Registration error using different projections.	Registration errors (µm)
Fig. S5b	BV erosion effect on neuron extraction.	BV erosion level (%), Tracking accuracy (%)
Fig. S6e	Spatial correlation of aligned neuron projections under z-axis displacements.	Correlation
Fig. S6f	Correlation of XY-shifts between different focal planes.	Reference focal plane (µm), +60 µm focal plane (µm)
Fig. S6g	Inter-session misalignment prediction across different focal planes.	Misalignment amplitude (µm)
Fig. S8b	Extraction precision from standard initialization (S1) vs. residual initialization (S2).	Precision
Fig. S8c	Overall extraction performance of CNMF-E batch vs. # of batches processed.	Number of sessions (#), F1-score (%)
Fig. S10c	Temporal similarities between components extracted at different focal shifts.	Neuron similarity score
Fig. S12d	Simulated and observed misalignment amplitude.	Misalignment amplitude (µm)
Fig. S12g	Clustering accuracy with UMAP in drifting conditions.	Clustering accuracy (%)
Fig. S13d	Reconstruction of segmented imaging data (5-minute segments).	Similarity with ground truth neurons (%)
Fig. S13e	Reconstruction of segmented imaging data (15-minute segments).	Similarity with ground truth neurons (%)
Fig. S14c	Mahalanobis distance of each component relative to the distribution from individually processed sessions.	Mahalanobis distance (D)
Fig. S14h	Percentage of neuropile classifications for CNMF-E and CaliAli.	Neuropile classifications (categorical variable)
Fig. S15b	Population vector distances of session pairs obtained across 99 days (undetected neurons have 0 activity).	Population vector distance
Fig. S15c	Session gap effect on PVD.	Session gap (days), PVD
Fig. S15d	PVD in non-aligned data.	Misalignment amplitude (µm), PVD

Extended Source Data

The following datasets are too large to be included in Source_data.xlsx and are provided as .mat files:

Data_4e.mat – Neuron Pair Distances Across Alignment Algorithms Utilizing Dual-Color Imaging as GT

Column	Description
Algorithm	Motion correction method used.
Projection	Type of projection applied (e.g., BV+Neurons, Neurons, BV, Filt. Mean).
m21 - m30	Mouse identities. Each column contains distances of GT neuron pairs (µm) in the corresponding mouse.

Data_6b-c.mat – Spatial Information Content of Place Cells

Column	Description
Activity	Average dF/F per neuron.
Info	Spatial information content.
Mouse_id	Mouse identifier (1-4).
Method	0 = Single session processing, 1 = CaliAli.
Context	Session number (1–4).

The table Standardized_Info_content contains the same variables, but Info is expressed as SD above chance (defined by random permutation).

Data_6h.pxp – Cumulative Probability of Place Field Correlation

This IgorPro file contains the cumulative probability distribution of place field correlation. The raw data used to generate this histogram is included in the source data file.

Data_7b-d.mat – Opto-Consistency Values for Neuron Tracking

Column	Description
Opto-consistency	1 if the neuron consistently responds to light across sessions, 0 otherwise.
Hyperparameter	Correlation threshold used for neuron matching.
Opto-threshold	Defines opto(+) neurons based on light response.
Method	Tracking approach used.

Neuron Extraction Data

Neuron extraction results from calcium imaging are provided as .mat files. Each file contains a table including some or all of the following variables:

a: [F×N double] Spatial components of extracted neurons, with F frames and N neurons, representing spatial activity over time.
c: [N×F double] Denoised temporal components, showing neuronal activity patterns after denoising.
cr: [N×F double] Raw temporal components, similar to c but without denoising.
d: [d1×d2] Video dimensions, specified as height (d1) × width (d2) in pixels.
Al: [d1×d2×3 uint8] Footprint projections, showing ground truth (GT) data, aligned extractions, and raw extractions.
m: [N×2 double] Temporal similarity of extracted neurons to GT neurons, measured using cosine similarity.
t: [100×5 table] Performance scores across different temporal similarity thresholds.
s: F1 score computed at a temporal similarity threshold of 0.8.

Included Datasets:

Dataset	Description
Fig_5m-p_neuron_data.mat	Simulated calcium imaging data evaluating SNR effects across sessions. Neurons were extracted using Single session processing or CaliAli (concatenated sessions), with high and low SNR neurons (defined by GT data) analyzed separately. The table contains four columns: Single session–High, Single session–Low, CaliAli–High, and CaliAli–Low.
Fig_6_neuron_data.mat	Place cell recordings. Cell array BH contains the behavioral data of mouse position, while T holds the extracted neuronal data.
Fig_7b-d_neuron_data.mat	Opto-tracking data. The Opto_imaging table includes rows for hyperparameters and columns for different extraction methods. The opt (cell array) contains light stimulation timings.
Fig_7f-h_neuron_data.mat	Opto-tracking data over a 4-week period using different methods. The opt (cell array) includes light stimulation timings.
Fig8_neuron_data.mat	Neuronal activity extracted over 99 days using different methods. Includes dateObjects, which records the timing of each session.
Fig_S5_neuron_data.mat	Simulated data evaluating tracking performance under varying blood vessel (BV) clarity. Columns represent levels of BV erosion, and rows represent simulation replicates. TCA contains neurons extracted using CaliAli, and TPC contains neurons extracted using Optimal Extraction.
FigS6_S10_neuron_data.mat	Simulated data evaluating tracking performance under focal plane displacements, with rows representing different focal plane shifts and columns corresponding to different tracking methodologies.
Fig_S9_neuron_data.mat	Simulated data where neuronal spatial components varied across sessions, with rows representing simulation replicates and columns corresponding to different extraction methods.
FigS12_neuron_data.mat	Simulated data under different conditions: Dense (dense neuronal distribution), Remapping (active neurons change across sessions), Ideal (non-dense, non-remapping distribution), and Drift (neurons gradually drift across conditions). Rows represent simulation replicates, and columns correspond to different extraction methods.

Code/Software

Code used to process these data sets are included in GitHub and in the Online Documentation of CaliAli