Models

class whobpyt.models.jansen_rit.jansen_rit.JansenRitModel(params: JansenRitParams, node_size=200, TRs_per_window=20, step_size=0.0001, output_size=64, tr=0.001, sc=array([[1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], ..., [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.]]), lm=array([[1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], ..., [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.]]), dist=array([[1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], ..., [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.]]), use_fit_gains=True, use_laplacian=True, use_fit_lfm=False)

A module for forward model (JansenRit) to simulate EEG signals

Methods

createIC(self, ver):	Creates the initial conditions for the model.
createDelayIC(self, ver):	Creates the initial conditions for the delays.
setModelParameters(self):	Sets the parameters of the model.
forward(input, noise_out, hx)	Forward pass for generating a number of EEG signals with current model parameters

__init__(params: JansenRitParams, node_size=200, TRs_per_window=20, step_size=0.0001, output_size=64, tr=0.001, sc=array([[1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], ..., [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.]]), lm=array([[1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], ..., [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.]]), dist=array([[1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], ..., [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.], [1., 1., 1., ..., 1., 1., 1.]]), use_fit_gains=True, use_laplacian=True, use_fit_lfm=False)

Parameters:

node_size: int

Number of ROIs

TRs_per_window: int # TODO: CHANGE THE NAME

Number of EEG signals to simulate

step_size: float

Integration step for forward model

output_sizeint

Number of EEG channels.

trfloat # TODO: CHANGE THE NAME TO sampling_rate

Sampling rate of the simulated EEG signals

sc: ndarray node_size x node_size float array

Structural connectivity

lm: ndarray float array

Leadfield matrix from source space to EEG space

dist: ndarray float array

Distance matrix

use_fit_gains: bool

Flag for fitting gains. 1: fit, 0: not fit

use_laplacian: bool

Flat for using laplacian. 1: yes, 0: no.

use_fit_lfm: bool

Flag for fitting the leadfield matrix. 1: fit, 0: not fit

params: ParamsJR

Model parameters object.

createDelayIC(ver, delays_max=500, state_lb=-0.5, state_ub=0.5)

Creates the initial conditions for the delays.

Parameters:

verint

Initial condition version. (in the JR model, the version is not used. It is just for consistency with other models)

Returns:

torch.Tensor

Tensor of shape (node_size, delays_max) with random values between state_lb and state_ub.

createIC(ver, state_lb=-0.5, state_ub=0.5)

Creates the initial conditions for the model.

Parameters:

verint # TODO: ADD MORE DESCRIPTION

Initial condition version. (in the JR model, the version is not used. It is just for consistency with other models)

Returns:

torch.Tensor

Tensor of shape (node_size, state_size) with random values between state_lb and state_ub.

forward(external, hx, hE)

This function carries out the forward Euler integration method for the JR neural mass model, with time delays, connection gains, and external inputs considered. Each population (pyramidal, excitatory, inhibitory) in the network is modeled as a nonlinear second order system. The function updates the state of each neural population and computes the EEG signals at each time step.

Parameters:

externaltorch.Tensor

Input tensor of shape (batch_size, num_ROIs) representing the input to the model.

hxOptional[torch.Tensor]

Optional tensor of shape (batch_size, state_size, num_ROIs) representing the initial hidden state.

hEOptional[torch.Tensor]

Optional tensor of shape (batch_size, num_ROIs, delays_max) representing the initial delays.

Returns:

next_statedict

Dictionary containing the updated current state, EEG signals, and the history of each population’s current and voltage at each time step.

hEtorch.Tensor

Tensor representing the updated history of the pyramidal population’s current.

info()

setModelParameters()

Sets the parameters of the model.

setModelSCParameters(small_constant=0.05)

Sets the parameters of the model.