Maith2021_ANNarchyBOLDmonitor

Source code of simulations and analyses from Maith, O., Dinkelbach, H. Ü., Baladron, J., Vitay, J., & Hamker, F. H. (2022). BOLD Monitoring in the Neural Simulator ANNarchy. Frontiers in Neuroinformatics.

Authors:

• Oliver Maith (oliver.maith@informatik.tu-chemnitz.de)
• Helge Ülo Dinkelbach (helge.dinkelbach@gmail.com)

Using the Scripts

Files

• dataRaw/
  • folder for generated data
• results/
  • folder structure for generated results
• srcAna/
  • BOLDfromDifferentSources_ANA.py
    • main analysis script, analyze data generated with BOLDfromDifferentSources in srcSim/simulations.py
    • loads data from dataRaw/, creates plots etc., saves results in results/BOLDfromDifferentSources/
    • in '__main__', define which plots are created
  • Fig_Balloon_model_data_ANA.py
    • analyze data generated with srcSim/Fig_Balloon_model_data.py
    • loads data from dataRaw/, saves results in results/Fig_Balloon_model_data/
  • BOLD_comp_time.py
    • analyze data generated with BOLDfromDifferentSources in srcSim/simulations.py (params["measure_time"]==True)
    • loads data from dataRaw/, saves results in results/Fig_comp_time/
  • extras.py
    • some extra functions available for importing
  • initialTestofBOLD_ANA.py
    • analyze data generated with initialTestofBOLD in srcSim/simulations.py
  • test.py
    • just placeholder for any testing
• srcSim/
  • Fig_Balloon_model_data.py
    • BOLD monitoring using a self-defined source signal (rectangular pulse)
  • extras.py
    • some extra functions available for importing
  • get_weightDist.py
    • for adjusting the weight distributions manually to obtain a good fit (visual comparison) between simulated and experimental distributions of post-synaptic potentials
    • generate 1000 RS and FS cortical neurons
    • send 1 input spike
    • compares (Figures) distribution of caused EPSPs in neurons with distribution of EPSPs from Song et al. (2005)
  • model.py
    • defintion of model (populations, projections)
    • different model versions (defined by params['optimizeRates'] in srcSim/parameters.py)
      • 'v1': used during optimization, model input = membrane currents, variable number of synapses for projections, self-defined input current distribution
      • 'v1post': version 'v1' but used after optimization (optimized parameters loaded)
      • 'v2': used during optimization, model input = Poisson neurons, fixed number of synapses for projections, scaled projection strengthes
      • 'v2post': version 'v2' but used after optimization (optimized parameters loaded), version used to generate results of the work
  • model_neuronmodels.py
    • defintion of neuron models and additional BOLD models
  • optimize_rates.py
    • two modes, optimize and test (script arguments)
    • works with model version 'v1'
    • optimize mode:
      • optimizes parameters (input current distribution and number of afferent synapses of all neurons)
      • goodness of fit: simulated mean firing rate distribution vs. distribution similar to the firing rate distributions shown in Buzsáki & Mizuseki (2014)
      • saves fitted parameters and loss
    • test mode:
      • load fitted params and prints loss
      • simulator optionally generates plots to also visually compare firing rate distributions
  • optimize_ratesv2.py
    • improved version of optimize_rates.py
    • works with model version 'v2'
    • here scaling factors of the projection strengthes are optimized
  • parameters.py
    • parameters which are used globally
  • run_parallel.sh
    • shell script for running python files
  • simulations.py
    • main simulation script
    • runs simulations with the cortical model and records signals of various BOLD monitors
    • data saved in dataRaw/
    • during simulation input is increased depending on the script arguments
      • argument 1, input_factor: defines how strong the increase of the firing rate of the input Poisson neurons is
      • argument 2, stimulus: defines length of simulation and timing of stimulus increase
      • argument 3, simID: simply a simulation ID for saving the data
  • test.py
    • python file for testing anything
  • test2.py
    • python file for testing anything

Results Pipelines

Description of the results generated in Python. For several figures, additional image processing software was used to create the final figures.

Result	analysis	simulation	comment
Figure 1	Fig_Balloon_model_data_ANA.py	Fig_Balloon_model_data.py	lineplots are generated individually in folder results/Fig_Balloon_model_data/
Figure 3		BOLD_fig3.py	plot is directly generated by the simulation script BOLD_fig3.py
Figure 4	BOLDfromDifferentSources_ANA.py (function rate_distribution)	simulations.py (standard arguments), get_weightDist.py	two plots are generated, weight distribution plot is directly generated by the simulation script get_weightDist.py
Figure 5	BOLDfromDifferentSources_ANA.py (function with_vs_without_normalization)	simulations.py X 1 Y	X = input_factor = 1--resting, 5--pulse; Y = simID = (0--39)
Figure 6	BOLDfromDifferntSources_ANA.py (function pulses_visualization)	simulations.py 5 1 Y	Y = simID = (0--39), lineplots are generated individually
Figure 7	BOLDfromDifferntSources_ANA.py (function make_sustained_input_plot)	simulations.py 1.2 3 Y	Y = simID = (0--39), lineplots are generated individually
Figure 8	BOLD_comp_time.py	simulations.py 1 0 1 1, simulations.py 1 0 1 2	vary number of neurons in parameters.py, params["measure_time"]==True is necessary
Figure S1	BOLDfromDifferntSources_ANA.py (function make_sustained_input_plot)	simulations.py 1.2 3 Y	Y = simID = (0--39), Figure 7 and S1 are generated together, lineplots are generated individually

Platforms

• GNU/Linux

Dependencies

• ANNarchy >= 4.7