Skip to content

jaxfne Suite: Multimodal Sensor Projections and EEG/MEG/EMM Proxy Bundle

A tutorial demonstrating separate projection pathways for scalp potential, magnetic field, and metabolic activity proxies.

Open in Colab

Learning Objectives

After completing this tutorial, you will understand:

  1. Multimodal Projections — how a single underlying population simulation drives multiple distinct macroscopic measurement streams.
  2. EEG-Proxy Projection — how to project laminar potential profiles to simulated scalp sensors.
  3. MEG-Proxy Projection — how to project oriented current dipoles to simulated magnetometer arrays.
  4. Metabolic EMM-Proxy — how to summarize activity costs as a separate metabolic energy proxy stream.
  5. Operator Separation — why separate sensor paths must remain distinct rather than merged into a single generic signal.

Biological/Computational Question

Question: How can we obtain separate, distinct macroscopic sensor readouts from a single neural population simulation, and how do their spatial projections differ?

Context: EEG registers volume-conducted scalp potentials. MEG registers magnetic fields induced by intracellular current dipoles. EMM-proxy summarizes metabolic activity cost. Projecting them via separate operator pathways ensures distinct signal profiles that represent complementary views of population dynamics.

Mathematical Glossary Flow

Here, we outline the foundational equations defining the distinct sensor paths:

1. EEG-Proxy Sensor Equation

  • Sensor definition: $\(Y_{\text{EEG}, s}(t) = \sum_{n=1}^{N} W^{\text{EEG}}_{sn} S_n(t)\)$
  • Definition of terms:
  • \(Y_{\text{EEG}, s}(t)\): Scalp potential readout at sensor \(s\) and time \(t\).
  • \(W^{\text{EEG}}_{sn}\): Leadfield weight from source \(n\) to sensor \(s\).
  • \(S_n(t)\): Source current from element \(n\).
  • \(N\): Number of source elements.
  • Worded description: Scalp potential is the linear leadfield projection of source currents onto scalp sensors.
  • Implementation location: fields.py

2. MEG-Proxy Sensor Equation

  • Sensor definition: $\(Y_{\text{MEG}, m}(t) = \sum_{n=1}^{N} W^{\text{MEG}}_{mn} S_n(t)\)$
  • Definition of terms:
  • \(Y_{\text{MEG}, m}(t)\): Magnetic field readout at sensor \(m\) and time \(t\).
  • \(W^{\text{MEG}}_{mn}\): Magnetic coupling weight from dipole \(n\) to sensor \(m\).
  • Worded description: Magnetic field is the linear projection of intracellular dipole current onto magnetometer sensors.
  • Implementation location: fields.py

3. EMM-Proxy Metabolism Equation

  • Energy definition: $\(Y_{\text{EMM}}(t) = \frac{1}{N} \sum_{n=1}^{N} |S_n(t)|\)$
  • Worded description: The metabolism estimate is the mean absolute source activity cost over time.
  • Implementation location: fields.py

Canonical Import

Every notebook script and library call standardizes to the canonical import:

import jaxfne as jtfne

All public APIs are called through the unified jtfne namespace.

Simulation Workflow

The tutorial walks through the standard jaxfne workflow:

Configuration -> construct -> simulate -> separate probes -> independent figures
  1. Configuration: Compose the population and register all target modalities via jtfne.Configuration().
  2. Construction: Build the model with jtfne.construct(cfg).
  3. Simulation: Run the vectorized simulation with jtfne.simulate(model, sim).
  4. Separate Probing: Compute EEG-proxy, MEG-proxy, and EMM-proxy readouts using separate operator calls.
  5. Separate Visualization: Plot distinct panel figures for each sensor modality and export the validation receipt.