Software, data & researcher profiles

AnestheSIM

An open, extensible patient simulator for closed-loop anesthesia, driven by VitalDB records.

AnestheSIM is my own open-source MATLAB/Simulink patient simulator for closed-loop anesthesia research and teaching. Data from the VitalDB open surgical database is read through a built-in layer that can be extended. Recorded or designed pump rates and end-tidal gas concentrations drive the following models, all with published parameters:

• covariate pharmacokinetic models of propofol, remifentanil and norepinephrine;
• three interchangeable BIS interaction models;
• an intravenous–inhalational model for sevoflurane and desflurane;
• mean arterial pressure and heart rate models.

The release includes cohort search, automatic inhaled-agent resolution, alarms, synthetic patients, unit tests, and the scripts that validate the simulator on VitalDB records.

Event-based fractional-order multivariable control of anesthesia driven by measured surgical events

Closed-loop anesthesia controllers are usually validated against made-up stimulation profiles: a fixed time course of assumed intensity, the same for every simulated patient. This dataset instead captures what actually perturbs an anesthetised patient, measured from clinical recordings. It comes with the code that extracts these events and the closed-loop results obtained by replaying them.

The events fall into three classes:

• 5,259 clinician actions, recorded on the target-controlled-infusion target channel;
• 16,170 physiological responses, detected with quality gating and thresholds set relative to each patient;
• 800 surgical protocol markers, as recorded.

Signals used are the bispectral index with its quality channel, propofol and remifentanil infusion, invasive arterial pressure and heart rate. Each case includes one hour of closed-loop simulation under fractional-order and integer-order event-based controllers.

• patients/Per-case signals, event table, closed-loop results, summary figure
• S0_index.csvCase list with demographics and event counts
• S1_patient_sheets.pdfPer-case figures, 181 patients
• S2_closedloop_metrics.csvPer-patient performance metrics
• code/Complete extraction pipeline with exact case list

Raw waveforms are not redistributed. The included case identifiers and code regenerate them from the public VitalDB API (Lee et al., Scientific Data 9:279, 2022).