Penalized matrix decomposition for denoising, compression, and improved demixing of functional imaging data

Calcium imaging has revolutionized systems neuroscience, providing the ability to image large neural populations with single-cell resolution. The resulting datasets are quite large (with scales of TB/hour in some cases), which has presented a barrier to routine open sharing of this data, slowing progress in reproducible research. State of the art methods for analyzing this data are based on non-negative matrix factorization (NMF); these approaches solve a non-convex optimization problem, and are highly effective when good initializations are available, but can break down e.g. in low-SNR settings where common initialization approaches fail. Here we introduce an improved approach to compressing and denoising functional imaging data. The method is based on a spatially-localized penalized matrix decomposition (PMD) of the data to separate (low-dimensional) signal from (temporally-uncorrelated) noise. This approach can be applied in parallel on local spatial patches and is therefore highly scalable, does not impose non-negativity constraints or require stringent identifiability assumptions (leading to significantly more robust results compared to NMF), and estimates all parameters directly from the data, so no hand-tuning is required. We have applied the method to a wide range of functional imaging data (including one-photon, two-photon, three-photon, widefield, somatic, axonal, dendritic, calcium, and voltage imaging datasets): in all cases, we observe ~2-4x increases in SNR and compression rates of 20-300x with minimal visible loss of signal, with no adjustment of hyperparameters; this in turn facilitates the process of demixing the observed activity into contributions from individual neurons. We focus on two challenging applications: dendritic calcium imaging data and voltage imaging data in the context of optogenetic stimulation. In both cases, we show that our new approach leads to faster and much more robust extraction of activity from the video data.

Download Full-text

Ultra-fast accurate reconstruction of spiking activity from calcium imaging data

Journal of Neurophysiology ◽

10.1152/jn.00934.2017 ◽

2018 ◽

Vol 119 (5) ◽

pp. 1863-1878 ◽

Cited By ~ 6

Author(s):

Vahid Rahmati ◽

Knut Kirmse ◽

Knut Holthoff ◽

Stefan J. Kiebel

Keyword(s):

Calcium Imaging ◽

Reconstruction Method ◽

Imaging Data ◽

Decay Kinetics ◽

Experimental Conditions ◽

Model Free ◽

Reconstruction Methods ◽

Wide Range ◽

Fast Processing ◽

Spiking Activity

Calcium imaging provides an indirect observation of the underlying neural dynamics and enables the functional analysis of neuronal populations. However, the recorded fluorescence traces are temporally smeared, thus making the reconstruction of exact spiking activity challenging. Most of the established methods to tackle this issue are limited in dealing with issues such as the variability in the kinetics of fluorescence transients, fast processing of long-term data, high firing rates, and measurement noise. We propose a novel, heuristic reconstruction method to overcome these limitations. By using both synthetic and experimental data, we demonstrate the four main features of this method: 1) it accurately reconstructs both isolated spikes and within-burst spikes, and the spike count per fluorescence transient, from a given noisy fluorescence trace; 2) it performs the reconstruction of a trace extracted from 1,000,000 frames in less than 2 s; 3) it adapts to transients with different rise and decay kinetics or amplitudes, both within and across single neurons; and 4) it has only one key parameter, which we will show can be set in a nearly automatic way to an approximately optimal value. Furthermore, we demonstrate the ability of the method to effectively correct for fast and rather complex, slowly varying drifts as frequently observed in in vivo data. NEW & NOTEWORTHY Reconstruction of spiking activities from calcium imaging data remains challenging. Most of the established reconstruction methods not only have limitations in adapting to systematic variations in the data and fast processing of large amounts of data, but their results also depend on the user’s experience. To overcome these limitations, we present a novel, heuristic model-free-type method that enables an ultra-fast, accurate, near-automatic reconstruction from data recorded under a wide range of experimental conditions.

Download Full-text

Automatic sorting system for large calcium imaging data

10.1101/215145 ◽

2017 ◽

Cited By ~ 4

Author(s):

Takashi Takekawa ◽

Hirotaka Asai ◽

Noriaki Ohkawa ◽

Masanori Nomoto ◽

Reiko Okubo-Suzuki ◽

...

Keyword(s):

Calcium Imaging ◽

Cell Activity ◽

Estimation Method ◽

Matrix Decomposition ◽

Real Data ◽

Imaging Data ◽

Initial State ◽

Sorting System ◽

The Stability ◽

Activity Estimation

SUMMARYIt has become possible to observe neural activity in freely moving animals via calcium imaging using a microscope, which could not be observed previously. However, it remains difficult to extract the dynamics of nerve cells from the recorded imaging data. In this study, we greatly improved the stability, and robustness of the cell activity estimation method via non-negative matrix decomposition with shrinkage estimation of the baseline. In addition, by improving the initial state of the iterative algorithm using a newly proposed method to extract the shape of the cell via image processing, a solution could be obtained with a small number of iterations. These methods were applied to artificial and real data, and their effectiveness was confirmed.

Download Full-text

Model-based decoupling of evoked and spontaneous neural activity in calcium imaging data

PLoS Computational Biology ◽

10.1371/journal.pcbi.1008330 ◽

2020 ◽

Vol 16 (11) ◽

pp. e1008330

Author(s):

Marcus A. Triplett ◽

Zac Pujic ◽

Biao Sun ◽

Lilach Avitan ◽

Geoffrey J. Goodhill

Keyword(s):

Spontaneous Activity ◽

Neural Activity ◽

Calcium Imaging ◽

Temporal Dynamics ◽

Population Level ◽

Imaging Data ◽

Evoked Activity ◽

Zebrafish Optic Tectum ◽

Low Dimensional ◽

Mouse Visual Cortex

The pattern of neural activity evoked by a stimulus can be substantially affected by ongoing spontaneous activity. Separating these two types of activity is particularly important for calcium imaging data given the slow temporal dynamics of calcium indicators. Here we present a statistical model that decouples stimulus-driven activity from low dimensional spontaneous activity in this case. The model identifies hidden factors giving rise to spontaneous activity while jointly estimating stimulus tuning properties that account for the confounding effects that these factors introduce. By applying our model to data from zebrafish optic tectum and mouse visual cortex, we obtain quantitative measurements of the extent that neurons in each case are driven by evoked activity, spontaneous activity, and their interaction. By not averaging away potentially important information encoded in spontaneous activity, this broadly applicable model brings new insight into population-level neural activity within single trials.

Download Full-text

Community-based benchmarking improves spike rate inference from two-photon calcium imaging data

10.1101/177956 ◽

2017 ◽

Cited By ~ 7

Author(s):

Philipp Berens ◽

Jeremy Freeman ◽

Thomas Deneux ◽

Nicolay Chenkov ◽

Thomas McColgan ◽

...

Keyword(s):

Neural Activity ◽

Calcium Imaging ◽

Indirect Measurement ◽

Generative Models ◽

Imaging Data ◽

Crowd Sourcing ◽

Two Photon ◽

Wide Range ◽

Improved Performance ◽

Highly Correlated

In recent years, two-photon calcium imaging has become a standard tool to probe the function of neural circuits and to study computations in neuronal populations. However, the acquired signal is only an indirect measurement of neural activity due to the comparatively slow dynamics of fluorescent calcium indicators. Different algorithms for estimating spike trains from noisy calcium measurements have been proposed in the past, but it is an open question how far performance can be improved. Here, we report the results of the spikefinder challenge, launched to catalyze the development of new spike inference algorithms through crowd-sourcing. We present ten of the submitted algorithms which show improved performance compared to previously evaluated methods. Interestingly, the top-performing algorithms are based on a wide range of principles from deep neural networks to generative models, yet provide highly correlated estimates of the neural activity. The competition shows that benchmark challenges can drive algorithmic developments in neuroscience.

Download Full-text

Online analysis of microendoscopic 1-photon calcium imaging data streams

PLoS Computational Biology ◽

10.1371/journal.pcbi.1008565 ◽

2021 ◽

Vol 17 (1) ◽

pp. e1008565

Author(s):

Johannes Friedrich ◽

Andrea Giovannucci ◽

Eftychios A. Pnevmatikakis

Keyword(s):

Neuronal Activity ◽

Calcium Imaging ◽

Computing Time ◽

Video Data ◽

Streaming Data ◽

Live Streaming ◽

Imaging Data ◽

Online Analysis ◽

Real Time Processing

In vivo calcium imaging through microendoscopic lenses enables imaging of neuronal populations deep within the brains of freely moving animals. Previously, a constrained matrix factorization approach (CNMF-E) has been suggested to extract single-neuronal activity from microendoscopic data. However, this approach relies on offline batch processing of the entire video data and is demanding both in terms of computing and memory requirements. These drawbacks prevent its applicability to the analysis of large datasets and closed-loop experimental settings. Here we address both issues by introducing two different online algorithms for extracting neuronal activity from streaming microendoscopic data. Our first algorithm, OnACID-E, presents an online adaptation of the CNMF-E algorithm, which dramatically reduces its memory and computation requirements. Our second algorithm proposes a convolution-based background model for microendoscopic data that enables even faster (real time) processing. Our approach is modular and can be combined with existing online motion artifact correction and activity deconvolution methods to provide a highly scalable pipeline for microendoscopic data analysis. We apply our algorithms on four previously published typical experimental datasets and show that they yield similar high-quality results as the popular offline approach, but outperform it with regard to computing time and memory requirements. They can be used instead of CNMF-E to process pre-recorded data with boosted speeds and dramatically reduced memory requirements. Further, they newly enable online analysis of live-streaming data even on a laptop.

Download Full-text

Virtual physiological human: training challenges

Philosophical Transactions of The Royal Society A Mathematical Physical and Engineering Sciences ◽

10.1098/rsta.2010.0082 ◽

2010 ◽

Vol 368 (1921) ◽

pp. 2841-2851 ◽

Cited By ~ 10

Author(s):

Patricia V. Lawford ◽

Andrew V. Narracott ◽

Keith McCormack ◽

Jesus Bisbal ◽

Carlos Martin ◽

...

Keyword(s):

Data Mining ◽

Knowledge Discovery ◽

Functional Imaging ◽

Imaging Data ◽

Database Development ◽

Wide Range ◽

Quantitative Changes ◽

Near Future ◽

Modelling Simulation

The virtual physiological human (VPH) initiative encompasses a wide range of activities, including structural and functional imaging, data mining, knowledge discovery tool and database development, biomedical modelling, simulation and visualization. The VPH community is developing from a multitude of relatively focused, but disparate, research endeavours into an integrated effort to bring together, develop and translate emerging technologies for application, from academia to industry and medicine. This process initially builds on the evolution of multi-disciplinary interactions and abilities, but addressing the challenges associated with the implementation of the VPH will require, in the very near future, a translation of quantitative changes into a new quality of highly trained multi-disciplinary personnel. Current strategies for undergraduate and on-the-job training may soon prove insufficient for this. The European Commission seventh framework VPH network of excellence is exploring this emerging need, and is developing a framework of novel training initiatives to address the predicted shortfall in suitably skilled VPH-aware professionals. This paper reports first steps in the implementation of a coherent VPH training portfolio.

Download Full-text

Localized semi-nonnegative matrix factorization (LocaNMF) of widefield calcium imaging data

10.1101/650093 ◽

2019 ◽

Cited By ~ 1

Author(s):

Shreya Saxena ◽

Ian Kinsella ◽

Simon Musall ◽

Sharon H. Kim ◽

Jozsef Meszaros ◽

...

Keyword(s):

Matrix Factorization ◽

Calcium Imaging ◽

Large Scale ◽

Nonnegative Matrix Factorization ◽

Nonnegative Matrix ◽

Brain Regions ◽

Video Data ◽

Imaging Data ◽

Dorsal Cortex ◽

Experimental Conditions

Widefield calcium imaging enables recording of large-scale neural activity across the mouse dorsal cortex. In order to examine the relationship of these neural signals to the resulting behavior, it is critical to demix the recordings into meaningful spatial and temporal components that can be mapped onto well-defined brain regions. However, no current tools satisfactorily extract the activity of the different brain regions in individual mice in a data-driven manner, while taking into account mouse-specific and preparation-specific differences. Here, we introduce Localized semi-Nonnegative Matrix Factorization (LocaNMF), a method that efficiently decomposes widefield video data and allows us to directly compare activity across multiple mice by outputting mouse-specific localized functional regions that are significantly more interpretable than more traditional decomposition techniques. Moreover, it provides a natural subspace to directly compare correlation maps and neural dynamics across different behaviors, mice, and experimental conditions, and enables identification of task- and movement-related brain regions.

Download Full-text

PatchWarp: Corrections of non-uniform image distortions in two-photon calcium imaging data by patchwork affine transformations

10.1101/2021.11.10.468164 ◽

2021 ◽

Author(s):

Ryoma Hattori ◽

Takaki Komiyama

Keyword(s):

Neural Activity ◽

Calcium Imaging ◽

Motion Correction ◽

Extended Period ◽

Affine Transformations ◽

Imaging Data ◽

Full Field ◽

Two Photon ◽

Automated Method ◽

Wide Range

Two-photon microscopy has been widely used to record the activity of populations of individual neurons at high spatial resolution in behaving animals. The ability to perform imaging for an extended period of time allows the investigation of activity changes associated with behavioral states and learning. However, imaging often accompanies shifts of the imaging field, including rapid (~100ms) translation and slow, spatially non-uniform distortion. To combat this issue and obtain a stable time series of the target structures, motion correction algorithms are commonly applied. However, typical motion correction algorithms are limited to full field translation of images and are unable to correct non-uniform distortions. Here, we developed a novel algorithm, PatchWarp, to robustly correct slow image distortion for calcium imaging data. PatchWarp is a two-step algorithm with rigid and non-rigid image registrations. To correct non-uniform image distortions, it splits the imaging field and estimates the best affine transformation matrix for each of the subfields. The distortion-corrected subfields are stitched together like a patchwork to reconstruct the distortion-corrected imaging field. We show that PatchWarp robustly corrects image distortions of calcium imaging data collected from various cortical areas through glass window or GRIN lens with a higher accuracy than existing non-rigid algorithms. Furthermore, it provides a fully automated method of registering images from different imaging sessions for longitudinal neural activity analyses. PatchWarp improves the quality of neural activity analyses and would be useful as a general approach to correct image distortions in a wide range of disciplines.

Download Full-text

MIN1PIPE: A Miniscope 1-photon-based Calcium Imaging Signal Extraction Pipeline

10.1101/311548 ◽

2018 ◽

Author(s):

Jinghao Lu ◽

Chunyuan Li ◽

Jonnathan Singh-Alvarado ◽

Zhe Charles Zhou ◽

Flavio Fröhlich ◽

...

Keyword(s):

Calcium Imaging ◽

Parameter Tuning ◽

Superior Performance ◽

Imaging Method ◽

Imaging Data ◽

Neural Signals ◽

Neural Activities ◽

Wide Range ◽

Photon Imaging

SUMMARYIn vivo calcium imaging using 1-photon based miniscope and microendoscopic lens enables studies of neural activities in freely behaving animals. However, the high and fluctuating background, the inevitable movements and distortions of imaging field, and the extensive spatial overlaps of fluorescent signals emitted from imaged neurons inherent in this 1-photon imaging method present major challenges for extracting neuronal signals reliably and automatically from the raw imaging data. Here we develop a unifying algorithm called MINiscope 1-photon imaging PIPEline (MIN1PIPE) that contains several standalone modules and can handle a wide range of imaging conditions and qualities with minimal parameter tuning, and automatically and accurately isolate spatially localized neural signals. We quantitatively compare MIN1PIPE with other existing partial methods using both synthetic and real datasets obtained from different animal models, and show that MIN1PIPE has a superior performance both in terms of efficiency and precision in analyzing noisy miniscope calcium imaging data.

Download Full-text

Online analysis of microendoscopic 1-photon calcium imaging data streams

10.1101/2020.01.31.929141 ◽

2020 ◽

Author(s):

Johannes Friedrich ◽

Andrea Giovannucci ◽

Eftychios A. Pnevmatikakis

Keyword(s):

Real Time ◽

Calcium Imaging ◽

Closed Loop ◽

Computing Time ◽

Video Data ◽

Imaging Data ◽

Real Time Processing ◽

Time Processing ◽

Freely Moving ◽

Deep Brain

AbstractIn-vivo calcium imaging through microendoscopic lenses enables imaging of neuronal populations deep within the brains of freely moving animals. Previously, a constrained matrix factorization approach (CNMF-E) has been suggested to extract single-neuronal activity from microendoscopic data. However, this approach relies on offline batch processing of the entire video data and is demanding both in terms of computing and memory requirements. These drawbacks prevent its applicability to the analysis of large datasets and closed-loop experimental settings. Here we address both issues by introducing two different online algorithms for extracting neuronal activity from streaming microendoscopic data. Our first algorithm presents an online adaptation of the CNMF-E algorithm, which dramatically reduces its memory and computation requirements. Our second algorithm proposes a convolution-based background model for microendoscopic data that enables even faster (real time) processing on GPU hardware. Our approach is modular and can be combined with existing online motion artifact correction and activity deconvolution methods to provide a highly scalable pipeline for microendoscopic data analysis. We apply our algorithms on two previously published typical experimental datasets and show that they yield similar high-quality results as the popular offline approach, but outperform it with regard to computing time and memory requirements.Author summaryCalcium imaging methods enable researchers to measure the activity of genetically-targeted large-scale neuronal subpopulations. Whereas previous methods required the specimen to be stable, e.g. anesthetized or head-fixed, new brain imaging techniques using microendoscopic lenses and miniaturized microscopes have enabled deep brain imaging in freely moving mice.However, the very large background fluctuations, the inevitable movements and distortions of imaging field, and the extensive spatial overlaps of fluorescent signals complicate the goal of efficiently extracting accurate estimates of neural activity from the observed video data. Further, current activity extraction methods are computationally expensive due to the complex background model and are typically applied to imaging data after the experiment is complete. Moreover, in some scenarios it is necessary to perform experiments in real-time and closed-loop – analyzing data on-the-fly to guide the next experimental steps or to control feedback –, and this calls for new methods for accurate real-time processing. Here we address both issues by adapting a popular extraction method to operate online and extend it to utilize GPU hardware that enables real time processing. Our algorithms yield similar high-quality results as the original offline approach, but outperform it with regard to computing time and memory requirements. Our results enable faster and scalable analysis, and open the door to new closed-loop experiments in deep brain areas and on freely-moving preparations.

Download Full-text