Related Experiment Video
Updated: Jul 11, 2025
Mapping Cortical Dynamics Using Simultaneous MEG/EEG and Anatomically-constrained Minimum-norm Estimates: an Auditory Attention Example
Published on: October 24, 2012
A recurrent Hopfield network for estimating meso-scale effective connectivity in MEG
Giorgio Gosti1, Edoardo Milanetti2, Viola Folli3
1Center for Life Nano- & Neuro-Science, Istituto Italiano di Tecnologia, Viale Regina Elena, 291, 00161, Rome, Italy; Soft and Living Matter Laboratory, Institute of Nanotechnology, Consiglio Nazionale delle Ricerche, Piazzale Aldo Moro, 5, 00185, Rome, Italy; Istituto di Scienze del Patrimonio Culturale, Sede di Roma, Consiglio Nazionale delle Ricerche, CNR-ISPC, Via Salaria km, 34900 Rome, Italy.
We developed a new model, the Recurrent Hopfield Mass Model (RHoMM), to understand brain communication patterns. RHoMM effectively estimates large-scale effective connectivity from MEG data, revealing insights into neural inhibition and excitation dynamics.
Area of Science:
- Neuroscience
- Computational Neuroscience
- Network Science
Background:
- The human connectome, mapping brain communication, is crucial in neuroscience.
- Understanding the generative models of neural inhibition/excitation remains a challenge.
- Existing models often require extensive hyperparameters and are less scalable.
Purpose of the Study:
- To introduce a novel generative model, the Recurrent Hopfield Mass Model (RHoMM), for estimating large-scale effective connectivity from MEG data.
- To address the challenge of identifying underlying patterns of neural inhibition and excitation.
- To provide a data-driven, scalable, and less hyperparameter-dependent alternative to existing biophysical models.
Main Methods:
- Developed the Recurrent Hopfield Mass Model (RHoMM) using a recurrent Hopfield neural network with asymmetric connections.
- Applied RHoMM to binarized Band Limited Power (BLP) dynamics derived from MEG data.
- Trained RHoMM using gradient descent minimization and validated its performance in capturing connectivity patterns and topological features.
Main Results:
- RHoMM demonstrated significant agreement between estimated effective connectivity and interregional BLP correlation, capturing individual variability.
- Simulated BLP correlation connectomes from RHoMM preserved key topological features, such as centrality, confirming model reliability.
- RHoMM showed a strong ability to predict MEG dynamics and effective connectivity.
Conclusions:
- RHoMM is a reliable and data-driven generative model for estimating large-scale effective connectivity from MEG.
- The model successfully captures individual variability in neural dynamics and preserves essential topological features of brain networks.
- RHoMM offers a scalable and computationally efficient approach for investigating neural inhibition/excitation dynamics across different spatial scales.
