Simulator¶

Bases: object

Class to add environment and usefull functions for simulation.

Parameters:

patient: Patient, optional: The virtual patient object. If None a random patient is generaed with random_generation_arg parameters.
random_generation_arg: dict, optional: Argument to pass to generate_random_patient_method. The default is empty.
tci_propo: str, optional: Type of TCI for Propofol. Can be either ‘Plasma’, ‘Effect_site’ or ‘none’. Defaults to ‘none’.
tci_remi: str, optional: Type of TCI for Remifentanil. Can be either ‘Plasma’, ‘Effect_site’ or ‘none’. Defaults to ‘none’.
disturbance_profil: str, optional: Type of disturbance profile to apply. See disturbance module for more details. The default is None.
noise: bool, optional: If True, add noise to the outputs of the patient model. The default is False.
save_signalsbool, optional: Save all internal variables at each sampling time in a data frame. The default is True.
bis_delay_maxfloat, optional: Maximum value of the BIS delay caused by Signal Quality Index (SQI) expressed in (s) according to the relationship proposed in [Wahlquist2025]. The default is 120 (s).
arg_disturbance: dict, optional: Additional argument to pass to the function compute_disturbances. The default is empty.
arg_tci_propo: dict, optional: Additional argument to pass to tci class initialization for propofol. The default is empty.
arg_tci_remi: dict, optional: Additional argument to pass to tci class initialization for remifentanil. The default is empty.
arg_tci_nore: dict, optional: Additional argument to pass to tci class initialization for norepinephrine. The default is empty.
arg_tci_atra: dict, optional: Additional argument to pass to tci class initialization for atracurium. The default is empty.

References

[Wahlquist2025]

Y. Wahlquist, et al. “Kalman filter soft sensor to handle signal quality loss in closed-loop controlled anesthesia” Biomedical Signal Processing and Control 104 (2025): 107506. doi: https://doi.org/10.1016/j.bspc.2025.107506

add_noise()¶

Add noise on MAP, HR and BIS.

All noise are considered white noise filtered by a second order transfert function:

For MAP, the standard deviation of the white noise is 30 and the filter is a second-order low-pass noise filter with damping ratio ξ=2 and cutoff frequency ω=0.01.
For HR, the standard deviation of the white noise is 17 and the filter is a second-order low-pass noise filter with damping ratio ξ=10 and cutoff frequency ω=0.02. In addition, the output is ceiled to the nearest integer.
For BIS, the standard deviation of the white noise is 35 and the filter is a second-order low-pass noise filter with damping ratio ξ=1 and cutoff frequency ω=0.04.

See identification details on this Notebook

Perform a simulation over multiple step times using the inputs profiles provided.

Parameters:

inputs_propoOptional[np.ndarray], optional: Infusion rates or TCI targets for propofol over time. The default is null.
inputs_remiOptional[np.ndarray], optional: Infusion rates or TCI targets for remifentnanil over time. The default is null.
inputs_noreOptional[np.ndarray], optional: Infusion rates or TCI targets for norepinephrine over time. The default is null.
inputs_atraOptional[np.ndarray], optional: Infusion rates or TCI targets for remifentanil over time. The default is null.
x0_propoOptional[np.array], optional: Initial state of the propofol PK model. The default is zeros.
x0_remiOptional[np.array], optional: Initial state of the remifentanil PK model. The default is zeros.
x0_noreOptional[np.array], optional: Initial state of the norepinephrine PK model. The default is zeros.
x0_atraOptional[np.array], optional: Initial state of the atracurium PK model. The default is zeros.
interpbool, optional: Whether to use zero-order-hold (False, the default) or linear (True) interpolation for the input array.

Returns:

pd.DataFrame: Dataframe including all the signals during the simulation.

generate_random_patient(distribution: str | None = 'uniform', patient_arg: dict | None = None)¶: Generate a random patient with characteristics following either a uniform distribution or a distribution fitted on VitalDB data.

init_dataframe()¶

Initilize the dataframe variable with the following columns:

‘Time’: Simulation time (s)
‘BIS’: Bispectral Index
‘SQI’: Signal Quality Index
‘LOC’: Loss of consciousness
‘TOL’: Tolerance level
‘TOF’: Train-of-four (%)
‘TPR’: Total eripheral resistance (mmHg min/ mL)
‘SV’: Stroke volume (ml)
‘HR’: Heart rate (beat/min)
‘MAP’: Mean Arterial Pressure (mmHg)
‘CO’: Cardiac Output (L/min)
‘u_propo’: Propofol infusion rate (mg/s)
‘u_remi’: Remifentanil infusion rate (µg/s)
‘u_nore’: Norepinephrine infusion rate (µg/s)
‘u_atra’: Atracurium infusion rate (µg/s)
‘x_propo_1’ to ‘x_propo_4’: States of the propofol PK model
‘x_remi_1’ to ‘x_remi_4’: States of the remifentanil PK model
‘x_nore’: State of the norepinephrine PK model
‘x_atra_1’ to ‘x_atra_4’: States of the atracurium PK model
‘blood_volume’: Blood volume (L)

if applicable TCI targets are also added:

‘target_propo’: Target concentration for Propofol (µg/ml)
‘target_remi’: Target concentration for Remifentanil (ng/ml)

one_step(input_propo: float = 0, input_remi: float = 0, input_nore: float = 0, input_atra: float = 0, blood_rate: float = 0, sqi: float = 100) → tuple[float, float, float, float]¶

Simulate one step of the patient model with given inputs.

If tci pumps are used, the inputs are the target concentrations. Otherwise, they are the infusion rates.

Parameters:

input_propofloat, optional: Infusion rate (mg/s) or target concentration (µg/ml) for Propofol. The default is 0.
input_remifloat, optional: Infusion rate (µg/s) or target concentration (ng/ml) for Remifentanil. The default is 0.
input_norefloat, optional: Infusion rate (µg/s) or target concentration (ng/ml) for Norepinephrine. The default is 0.
input_atracuriumfloat, optional: Infusion rate (mg/s) or target concentration (µg/ml) for Atracurium. The default is 0.
blood_ratefloat, optional: Fluid rates from blood volume (mL/min), negative is bleeding while positive is a transfusion.
sqi: float, optional: Signal Quality Index of the BIS signal. It affects the BIS delay (expressed in seconds) according to the relationship proposed in [Wahlquist2025]: \(bis\_delay = bis\_delay\_max * (1 - \frac{sqi}{100})\). The default is 100.

Returns:

tuple[float, float, float, float]: BIS, MAP, HR, TOF values after one step of simulation.

References

[Wahlquist2025]

save_data(inputs: list = None)¶: Save all current internal variables as a new line in self.dataframe.