Cortical network structure mediates response to brain stimulation

A method for treating a neural disorder of a subject, comprising:

• delivering an input stimulus to an area of a brain, via a cortical implant attached to the subject;
• in response to delivering the input stimulus, generating neural response signals in the brain;
• recording the neural signals from a surface of the brain or a depth of the brain, wherein the neural signals from the brain represent brain activity;
• determining a predicted outcome, wherein the predicted outcome is based on a set of data derived from a nonlinear model that combines;
• protocol features that are brain agnostic, and
• network features that are based on interconnections between neural nodes of the brain, wherein the network features are computed from local field potential (LFP) time series recorded during a resting state block;
• updating the input stimulus into an evolved therapeutic input stimulus dynamically over time based on the predicted outcome,
• obtaining a coherence tensor by time-concatenation of coherence matrices across windows in the resting-state block that precedes the evolved therapeutic input stimulus for which functional connectivity changes (FCC) are predicted;
• deriving the functional connectivity changes from the coherence tensor;
• assessing importance of the network features by evaluating a difference in a predictive outcome between the nonlinear model and a model fit, wherein the model fit is the nonlinear model without one or more network features of interest;
• based on the predicted outcome, delivering the evolved therapeutic input stimulus to the subject through the cortical implant;
• and
• in response to delivering the evolved therapeutic input stimulus, generating evolved neural response signals in the brain.