Version: 5.0
Release Date: November 2025
Authors: Ronald Garcia Reyes, Ariosky Areaces Gonzales, Pedro A. Valdes–Sosa, Xi-AlphaNET Research Team

Xi-AlphaNET is a parametric, multivariate generative framework designed to map lifespan trajectories of EEG source spectral dynamics at high spatial resolution across large populations. Grounded in biophysical modeling, the framework jointly estimates source-level spectral parameters, effective connectivity, and interareal conduction delays while remaining computationally scalable. In its current formulation, Xi-AlphaNET explicitly models aperiodic (ξ) activity and the alpha rhythm (α) as anatomically constrained stochastic processes, enabling characterization of their distinct spatiotemporal organization across the cortex. Applied to large, harmonized EEG cohorts, the model reveals systematic age-dependent effects, including nonlinear inverted-U trajectories of spectral parameters and global conduction delays, as well as an isotropic spatial organization of aperiodic activity. Future releases will extend this framework to incorporate additional spectral components, providing a more comprehensive generative description of cortical electrophysiological dynamics across the lifespan.

After processing each subject, Xi-AlphaNET produces a standardized set of output files stored in the designated results directory. These files contain source-level spectral estimates, connectivity measures, and model-specific parameters:

• Supported input formats include raw EEG recordings (.PLG, .EDF, .MAT, .set) and, optionally, precomputed cross-spectral data stored in .mat format. Plain-text (.txt) input will be supported in future releases.
• For optimal results, EEG preprocessing should be performed on raw EEG signals prior to any spectral estimation. Artifact rejection, bad-channel identification, re-referencing, filtering, and epoch selection should be carried out using established EEGLAB functions.
• The use of precomputed cross-spectra is supported for advanced users; however, preprocessing at the cross-spectral level is not recommended, as it limits artifact control and channel-level quality assessment.
• Organize the dataset such that each subject is contained in a separate subfolder.
• Run Update_data.m from the /utils/ directory to standardize the dataset structure:
  • Define input_folder as the directory containing the original EEG data.
  • Define output_folder as the directory where the structured dataset will be created.

• Run get_participants_info.m.
• This step generates a Participant.json file containing participant metadata (e.g., age, sex).
• The input path must correspond to the structured dataset created in Step 1.

• Launch Xi-AlphaNET using one of the following methods:
  • MATLAB command window: xialphanet

• Upon startup, Xi-AlphaNET automatically checks for previously saved analyses.
• Select:
  • Load to open existing results in the interface.
  • Close to start a new analysis with a fresh dataset.

• Dataset Name: Specify a project identifier.
• Description: Optional textual description of the dataset.
• Input Path: Directory containing the EEG or cross-spectral data (from Step 1).
• Reference File: Select an example EEG file (.PLG, .EDF, .MAT, or .set) to define the expected data structure.
• Participant File: Load the Participant.json file generated in Step 2.
• Data Type: Select either raw EEG signals or cross-spectral data.

• Output Path: Directory where all results will be stored.
• Temporary Path: Location used for intermediate files during computation (local by default).
• Parallel Computation: Enabled by default and recommended for large datasets; can be disabled if required.
• Save Mode: Option to store extended outputs, such as intermediate spectral quantities.

• By default, Xi-AlphaNET uses the HCP-MMP1 cortical parcellation, which defines the spatial resolution of the anatomical connectivity and conduction delay matrices.
• Users may optionally disable the default anatomy and provide custom anatomical inputs:
  • Cortical surface
  • Lead field
  • Channel definition file
  • Anatomical connectivity matrix
  • Tract length information
  • Conduction delay matrix
• Important: All anatomical inputs must be defined in the same spatial reference frame and be dimensionally consistent.

• Click Apply to save the current configuration.
• Click Run to start model estimation.

Note: Runtime depends on dataset size, parameter configuration, and available computational resources. The software scales efficiently on multi-core systems using parallel computation.

• Upon completion, the processed dataset appears in the Datasets panel.
• Expanding a dataset displays:
  • Left panel: List of processed subjects.
  • Right panel: Available visualizations (e.g., alpha power maps).

• Selecting a subject enables visualization of:
  • Cortical distributions of α and ξ power
  • Effective connectivity and conduction delay matrices
  • Source-level power spectral densities
  • Spatial maps of spectral parameters, including peak alpha frequency, bandwidth, and spectral exponent

• Primary reference:
  Lifespan Development of EEG Alpha and Aperiodic Component Sources is Shaped by the Connectome and Axonal Delays
• Support contact:
  ronaldo@neuroinformatics-collaboratory.org