LFP

[WeinaJi]

Future Work

This document outlines planned future enhancements and features for the Local Field Potential (LFP) functionality.

Multiple Weights Files Support

Currently, the "Run" block only supports a single weights file. Future implementation will:

• Enable multiple weights files to be specified
• Define how multiple files interact and are processed
• Update the configuration schema accordingly

Electrode Subset Selection

Implement functionality to allow users to select specific subsets of electrodes instead of using all available electrodes:

• Direct selection via electrode names
• Selection through metadata filtering (e.g., spatial coordinates: all electrodes where "x" < 500)
• Define syntax for subset specification in configuration

Weights Calculation from Circuit

Enable direct weights calculation from circuit information:

• Remove dependency on running simulation with single timestep (see https://github.com/BlueBrain/BlueRecording)

Circuit Building Integration

Automate electrode grid generation during circuit building:

• Generate standard electrode configurations
• Integrate with existing circuit building workflow
• Provide customization options for grid parameters

Advanced Configuration in SONATA

Define and implement advanced feature configuration in SONATA:

• Specify syntax for electrode subset selection
• Document configuration options
• Possible example
  

Libsonata Integration

Implement electrode file reading capabilities in libsonata:

• Move scaling factor reading from neurodamus to libsonata

LFP Report Output Options

Implement flexibility in LFP report generation to manage storage requirements:

• Enable choice between:
  • Full cell traces (current behavior)
  • Summed LFP only (sum over cell traces)
• Configuration options:
  • Via SONATA config using 'sections' key:
    • "sections": "all" -> report all node_ids
    • "sections": "soma" -> sum across cells
• Technical considerations:
  • Implementation uses cellular reports (ignoring node_ids). Basically extracellular reporting.
  • Supports storage optimization (e.g., reducing 12 TB to manageable size for 50s recording)
• Use case example:
  • 32 channel sites
  • 20 KHz sampling frequency
  • Support for long duration recordings (1+ hour)
• Alpha implementation
  • NEURON branch: https://github.com/neuronsimulator/nrn/tree/jblanco/lfp_sum_cells
  • Spack branch: libsonata-report, libsonata, py-libsonata, py-neurodamus, neuron: LFP features spack#1953
  • Implements core functionality for LFP summing over cells
  • Limitation: Number of ranks should be smaller than node_ids in LFP report target

Useful tools

Script for offline calculation of lfp from full compartment reports that uses weights files:
lfp_calc_updated.py.zip