FocusStack and StimServer

New, open-source toolboxes for visual stimulation and analysis of two-photon imaging data

1st December, 2014
Multiple-trial GCAMP6m calcium responses to a natural movie stimulus.

Figure 1: Overview of an experiment setup.

Figure 1: Overview of an experiment setup.

StimServer is used to generate and present visual stimuli to an animal, under the control of a calcium imaging system, over a network link. Data is stored in a binary format.

Figure 2: Importing and using raw two-photon imaging stacks.

Figure 2: Importing and using raw two-photon imaging stacks.

A FocusStack object is created in Matlab, to access several sequential data blocks as a single concatenated stack. The FocusStack object "fs" can now be accessed as a Matlab tensor, and passed into Matlab functions. Extracting a single frame is as simple as referencing a Matlab tensor. Extracting the response trace through time of a single pixel is equally simple.

Figure 3: Frame meta-data

Figure 3: Frame meta-data

Stimulus information and other meta-data is associated with each frame. Using "FocusStack.FrameStimulusInfo", the stimulus meta-data associated with each frame can be accessed (meta-data is listed for the frame indicated by the arrow).

Figure 4: Extracting orientation-tuned responses

Figure 4: Extracting orientation-tuned responses

Measuring preferred orientation using drifting gratings (top). Centre: single-trial single-neuron responses to drifting high-contrast gratings. Bottom: The trial-averaged direction tuning curve for the neuron shown above. Data provided by M. Roth.

Two-photon calcium imaging of neuronal responses is an increasingly accessible technology for probing population responses in cortex at single cell resolution, and with reasonable and improving temporal resolution. However, analysis of two-photon data is usually performed using ad-hoc solutions. To date, no publicly available software exists for straightforward analysis of stimulus-triggered two-photon imaging experiments, while incorporating information about stimulus triggers. In addition, the increasing data rates of two-photon acquisition systems imply increasing cost of computing hardware required for in-memory analysis.

FocusStack is a new, open-source, Matlab-based two-photon analysis toolchain designed to process large two-photon imaging stacks with only a small memory footprint. This makes analysis possible on comparatively cheap consumer hardware. FocusStack is designed to integrate with StimServer, a new open-source Matlab-based server for visual stimulus generation and presentation which can be controlled remotely over TCP or UDP network links.

Accessing the toolboxes

Both toolboxes are available as open source repositories:
FocusStack: Wiki; GIT repository; 20141118 release (.zip)
StimServer: Wiki; GIT repository; 20141118 release (.zip)

Getting started with FocusStack

Working with FocusStack

The design goal of FocusStack was to provide a simple, extensible toolchain to assist experiments using two-photon calcium imaging of neuronal responses; accessible to those with little programming experience, but powerful enough to automate most low-level analysis of calcium response stacks.

Due to the real-time requirements of both visual stimulation and acquisition of two-photon imaging data, these tasks are usually performed on separate dedicated computing systems. StimServer is controlled over a TCP or UDP network link, to trigger stimulus presentation and sequencing (Figure 1). Two-photon acquisition occurs using the software appropriate for the experimental equipment used, and stores the resulting imaging stacks as binary data files on disk. Ideally, meta-data about the stack — stimulus identity and random sequencing, stack resolution, information about the acquisition system, etc. — are stored with the stack data files in a file header or a “side-car” meta-data file.

Binary stack data files are analysed in Matlab, by using FocusStack to map several stack files to a single FocusStack object (Figure 2). This object appears as a simple Matlab tensor, with frames, channels and single pixels accessed using standard Matlab referencing. Since FocusStack objects can be accessed as Matlab tensors, many existing Matlab analysis functions that expect tensors can seamlessly be passed FocusStack objects without modification.

However, FocusStack objects are aware of stimulus timing and sequencing, provide services for stack alignment, provide support for assigning baseline fluorescence distributions, and have simple helper functions to perform derandomization and segmentation of stack data. Although a "FocusStack" object can be accessed just like a Matlab tensor, each frame and pixel in the stack has many items of meta-data associated with it (Figure 3). Stack-global information such as the stack resolution, or stimulus-specific meta data such as the stimulus presented during the acquisition of a given frame, is accessed using the FrameStimulusInfo method.

Calcium trace extraction and stimulus derandomisation

In any good experiment design, stimulus presentation order is randomized. During analysis of the acquired time series data, stimulus segmentation and derandomization therefore becomes an important but fiddly task. Our solution is to store the stimulus presentation order with the stack, along with information about stimulus duration, “blank” stimuli, and periods of stimulus presentation during which analysis of the calcium signals should be performed (see Figure 3).

Extracting calcium response time-series from a FocusStack object is accomplished using the ExtractRegionResponses function. This workhorse function transparently performs stimulus derandomization, simultaneously averages and extracts responses from a number of arbitrary ROIs in the stack, segments the stack into single-trial per-neuron traces and returns estimated responses for each stimulus and each trial.

The following code illustrates how complete calcium traces, as well as derandomised single-trial stimulus responses, can be extracted from experimental data in five lines of Matlab code. Note that information concerning stimulus presentation sequence is usually loaded along with the raw imaging data, in which assigning cvnSequenceIDs manually is unnecessary.

% - Create a FocusStack object
fs = FocusStack({'block1.bin', 'block2.bin'});

% - Assign stimulus presentation order
%   (usually loaded from stack data file)
fs.cvnSequenceIDs = {[2 3 1 4], [4 2 3 1]};

% - Assign stimulus timing information
fs.vtStimulusDurations = [4 4 4 4];
fs.mtStimulusUseTimes = [0 4; 2 4; 2 4; 2 4];

% - Segment, derandomize and extract
%   calcium responses and traces
[vfBlankStds, mfStimMeanResponses, mfStimStds, ...
   mfRegionTraces, tfTrialResponses, tnFramesInSample, ...
   cvfTrialTraces, mfRawRegionTraces] = ...
      ExtractRegionResponses(fs, sRegions, nBlankStimID);

Analysis example — direction tuning

Figure 4 illustrates the extraction of orientation and direction tuning from multiple-trial calcium responses, evoked by drifting high-contrast grating stimuli presented to an anaesthetised mouse (using StimServer). This analysis is included in the ???example code linked above.

Publications

This work was published in Frontiers in Neuroinformatics: DR Muir and BM Kampa. 2015. FocusStack and StimServer: A new open source MATLAB toolchain for visual stimulation and analysis of two-photon calcium neuronal imaging data, Frontiers in Neuroinformatics 8 85. Please cite this publication in lieu of thanks, if you use either toolbox.

Acknowledgements

The authors would like to express effusive thanks to M. Roth for acquiring the experimental data analysed here, and for making that data available for download. We gratefully acknowledge the contributions of the users of FocusStack and in locating and fixing bugs, and those who also contributed code to the toolboxes. In particular, we would like to thank M. Roth, P. Molina-Luna and A. Keller for their contributions.

Funding

This work was supported by the Novartis Foundation (grant to DRM), Velux Stifting (grant to DRM), the Swiss National Science Foundation (Grant Nr. 31-120480 to BMK), and the EU-FP7 program (BrainScales project 269921 to BMK).