Fig. 4: Multimodal connectomics model.
From: Longitudinal excitation-inhibition balance altered by sex and APOE-ε4
A structural connectome (A) is generated from DWI, with positive coefficients (Cij) only, and is used to regularize the Ising model (C). Preprocessed fMRI time-series data is binarized (B) and embedded (through a maximum pseudolikelihood function) onto the Ising model as interacting units of brain regions (“spins”). The optimal Ising model embedding results in functionally-informed structural connectome (D), called the resting-state structural connectome (rsSC), with positive and negative coefficients, which can be inferred as globally excitatory or inhibitory connections, respectively. From the rsSC, we can infer the balance of excitation-inhibition through an excitation-inhibition ratio (EIR).
