Connectivity Analysis

To quantify the efficiency of connection graph, we can use MST ratio (MST weigh / total weight).

Reference

https://onlinelibrary.wiley.com/doi/epdf/10.1002/hbm.24014

While working on #32, I've noticed an error in examples/post_processing/spike_detection.py that appears to be caused by a stale import from miv.io import load_data. I suspect that interface in miv.io was recently updated as it does not contain a load_data function. Full stack trace can be viewed here

Reference

• Spike detection for large neural populations using high density multielectrode arrays

Could be useful:

• https://elife-asu.github.io/PyInform/

Milestone v0.3

New Features

• #83 @Gauravu2
  • Partial Information Decomposition Lattice (PID)
    • [Timme et al., 2016]
    • [Luppi et al., 2022]
    • Newman et al., 2022
• Spatiotemporal coherence #62 @Gauravu2
• Data Encoding/Decoding to Spike (Coding)
  • Temporal
    • (Audio) Lyon's Ear Model @skim0119 #91
    • (Signal) Bens Spiker Algorithms Encoder @skim0119 #91
    • (Temporal Contrast) Threshold-based representation (TBR), Step-Forward (SF), Moving Windows (MW) @skim0119
    • (Signal) Hough Spiker Algorithm (HSA/mHSA)
    • (Audio) Spectrogram (scipy)
    • (Signal) LIF coding
  • Spatial
    • Gaussian Receptor Field
    • Bohte GRF
    • First-Spike (TTFS)
    • LIF coding
    • Rate-based encoding
  • Difference
    • Signal to noise ratio (SNR)
    • Root mean square error (RMSE)
    • Coefficient of regression
• "Memory" Quantification
  • Fading Memory
  • Edge-of-Chaos
  • Separation and Generalization
• TTL Digital input signal readout #99 @skim0119

Additional Functionality

Here is the list of functions we would like to implement using our code. Some could be accomplished by having "how-to" documentation, some might need additional implementations on the modules. Feel free to make a separate issue and propose ideas if you want to take a shot.

• Adaptive filtering
• #68 @jihugo
• Open Sound Control (OSC) plugin

Data Organization

• #90 @iraikov
• Storage for experiment/simulation dataset
• ASDF to SpikeTrains #77 @skim0119

Documentation

• Logo
• About page

Test Coverage

Others

• #37
• Semi-online processing for long recording

Following redundancies associated with Poetry can be eliminated by following GazzolaLab/PyElastica#141

1. The coverage info in pyproject.toml is redundant and not utilized currently in the presence of .coveragerc
2. The sphinx-ci can be much more compacted following GazzolaLab/PyElastica#141

Here is the reference : https://neuraldatascience.io/6-single_unit/intro_spike_trains.html

Documentation on how to use cross-correlation analysis using Elephant.

Neural data has been traditionally analyzed either temporally (spectral analysis) or spatially (connectivity measures). DMD can be used to extract spatiotemporal modes, hence combining the benefits of aforementioned techniques in one analysis and measure coherence between electrodes.
It is widely used for fluid mechanics problems, but not as common in neuroscience. However, some authors have started exploring DMD's for neuroscience [1].

References:

1. https://www.sciencedirect.com/science/article/abs/pii/S0165027015003829
2. https://www.annualreviews.org/doi/pdf/10.1146/annurev-control-071020-010108#article-denial

TODO

• implementation
• doc
• test (Will be finished with integrated test and documentation (dataset->statistics guide)

Currently, the coverage of MiV-OS sits at ~40%, indicating room for improvement. Below I will dynamically populate a checklist, listing the parts that are majorly untested, which will be resolved by adding tests step by step:

• miv/visualization/waveform.py
• miv/visualization/fft_domain.py
• miv/visualization/event.py

and so on...

Sometimes neurons fire at a higher frequency for a short period of time. Such events are called bursts and are characterized by burst firing rate and length. Bursting is usually taken as an indication of culture's maturation and robustness. More details could be found here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4725104/

Functions

• Load and unload the raw signal
• Keep the meta-data structure
• Iterable access
• Mask channels, prune the datatree
• Export(will be added in functional module)
• Interface (neo provides interface with other data format)

Note

• User should not be able to alter the raw data

Milestone v0.2

Not all features will be included for v0.2, and the remaining will be pushed to the next release.
If an external package is already available, a simple wrapper (if needed) with detailed documentation on how to use that package could be sufficient.

New Features

For each category, the sublist is in the order of priority.

• Data Conversion
  • List of SpikeTrains to scipy sparse matrix format @skim0119 (Included in elephant.conversion)
• Connectivity
  • #41 @Gauravu2
  • #58 @Gauravu2
  • coherence #51 @Gauravu2
  • granger causality #51 @Gauravu2
  • #70
  • #69
  • directed transfer information #58 @Gauravu2
  • mutual information #58 @Gauravu2
  • misc information theory #58 @Gauravu2
    • shannon-entropy
    • block entropy
    • entropy rate
    • active information
    • relative entropy
    • conditional entropy
    • transfer entropy
• Criticality
  • #40 @frthjf
  • Example Dataset (7bb3486) @skim0119
• Spike Classification / Quantification
  • Wavelet/PCA characterization
  • Correlation Matrix 74f7a47 @jihugo
• Statistics / Analysis
  • #43 @armantekinalp
  • #46 @skim0119
  • #49 @armantekinalp
  • #54 @Gauravu2
  • Fano Factor #65 @Gauravu2
  • Tuning curves
• Visualization
  • MEA geometry visualization (#24) @skim0119
  • #47 @Gauravu2
  • #59 @armantekinalp

Additional Functions

Here is the list of functions we would like to implement using our code. Some could be accomplished by having "how-to" documentation, some might need additional implementations on the modules. Feel free to make a separate issue and propose ideas if you want to take a shot.

• #36 @jihugo

Data Organization

• Simple practice dataset #38 #39
• External data installation utilities #39
  • In-line download for Colab or remote-server
  • Hash check for large file
  • Archive extraction

Documentation

Coverage

• #64

Others

• Plugin interface for MiV-Simulator

Need to setup:

• pypi
• travis (@armantekinalp are we gonna use travis or just use github-action)
• github-action (#5)
• read-the-doc (#3)
• pepy
• codecov
• gitter(maybe)

@skim0119 There is an offset in the event timestamps saved with OpenEphys digital TTL input and the timestamps of neurons firing observed in cultures. The offset is non-uniform across days and samples, hence, probably is some precision issue.

A similar issue with digital TTL input has been discussed here: open-ephys/GUI#302

Matlab asdf.mat format converter.

The package lyon on pypi is outdated, so I'm using github-link as a dependency, but pypi cannot handle.

Note

The data module has a masking feature, but the channel indexing needs to be preserved for later usage. We can have an explicit index map stored separately, while re-define getter returns a masked signal.

Spiketrain statistics

Begg's group commented that subtracting the median across all electrodes at each time step might help in getting rid of the false spikes as they occur simultaneously. Then the conditioned signal can be passed to spike detection and checked if we get lesser false spikes.

Setup pyroject.toml

poetry

• dependencies (requirements.txt)
• transfer setup.py

• black (21.12b0)
• autoflake8
• mypy
• tox

https://github.com/brian-team/brian2hears

The Fano factor, defined as the variance-to-mean ratio of spike counts in a time window, is often used to measure the variability of neuronal spike trains.

Currently, pypi deployment is blocked by non-registered repositories.

Note

Blocked by #97

Channel axis for readout signal is in column.

Note

A feature should be included in miv.io.Data which automatically determines which channels to mask. Channels with no spike (white noise) or faulty spike (non-neural spike) should be masked.

Possible syntax

>>> data = Data(data_path)
>>> spontaneous_data = Data(spontaneous_data_path)
>>> data.mask_channel(spontaneous_data)

Need to include the following:

• Bivariate Partial Information Decomposition (Unique Information, redundant information, synergy)
• Multivariate Partial Information Decomposition (,,)
• Multivariate Conditional Entropy (Decomposition included)
• Multivariate Transfer Entropy (Decomposition included)

Refer the following for more info:
https://arxiv.org/pdf/1004.2515.pdf
https://arxiv.org/pdf/1102.1507.pdf
https://pwollstadt.github.io/IDTxl/html/idtxl_process_analysis.html#bivariate-partial-information-decomposition-pid

Multi-channel signal visualization

• Output video file that played the recorded signal
• The layout of the plot correspond to the MEA channels.

Note

• You can implement in miv/visualization.

Candidates

• quantities: scientific units
• neo: spike data structure
• elephant: spike data processing (rasterize, )

Visualization tool

• ephyviewer
• viziphant

Artificial spike trains generated using Brian2 can be used to check our algorithms and also extends the user base of MiV OS

Update on basic set of visualization and reference documentation.

Hello,
This is a request for the creation of a small recording data set with valid signals that could be used to drive the development of an MiV native data format and associated data structures and algorithms. Thanks!

Tasks

Graph visualization given multi-channel patterns.

• graphviz

Loading MCS generated .h5 file can expand the utility of MiV-OS.
Their data format is available here: https://www.multichannelsystems.com/sites/multichannelsystems.com/files/documents/manuals/HDF5%20MCS%20Raw%20Data%20Definition.pdf

TODO

• Overall signal filter module
• Protocol to interface with data
• Unit-test
• Documentation
  • API
  • Usage

Update v0.1.0

We are targeting below set of post-analysis tools for first release.

Data Readout

• #11

Post-processing

• Filter
  • #10
• Detection
  • #13
  • clustering - guide on using SciPy PCA
• Statistics
  • #15
• Connectivity
  • transfer entropy
  • directed information (pushed to 0.2.0)
  • unitary event
  • SPADE (elephant)
  • CAD (elephant)
• Visualization
  • Spike cutout and guide #16
  • Rasterplot + rate (viziphant)
  • Connectivty graph (graphviz)

Future

• Wavelet clustering
• Geometry of MEA module

TODO

• basic spike detection
• protocol
• test (left functionality test and integration test for future work)
• doc