Emulation Notes¶

This document summarizes the best‑effort references used for the emulator models in IINTS‑AF. These emulators are approximations, intended for research comparisons — not clinical use.

What’s Implemented¶

Medtronic MiniMed 780G: src/iints/emulation/medtronic_780g.py
Tandem t:slim X2 Control‑IQ: src/iints/emulation/tandem_controliq.py
Omnipod 5: src/iints/emulation/omnipod_5.py

Each emulator exposes emulate_decision(...) and includes get_sources() citations.

Methodology (High‑Level)¶

Parameter extraction from public clinical papers, FDA summaries, and user manuals.
Behavioral modeling of typical decision patterns (auto‑basal, correction bolus, suspend logic).
Safety constraints encoded via SafetyLimits.

Limitations - Algorithms are proprietary; these are informed approximations. - Parameter ranges may differ by region / firmware. - Emulators represent a “typical” behavior, not individual personalization.

References (Examples)¶

Medtronic 780G¶

Bergenstal et al. (2020), NEJM: DOI 10.1056/NEJMoa2003479
FDA 510(k) K193510
Medtronic 780G User Guide (public manual)

Tandem Control‑IQ¶

Brown et al. (2019), Diabetes Technology & Therapeutics: DOI 10.1089/dia.2019.0226
FDA 510(k) K191289
Control‑IQ User Guide

Omnipod 5¶

Omnipod 5 Pivotal / ASSERT trial publications
FDA 510(k) K203467
Omnipod 5 User Guide

Example Usage¶

from iints.emulation.medtronic_780g import Medtronic780GEmulator

emulator = Medtronic780GEmulator()

# Example inputs
decision = emulator.emulate_decision(
    glucose=180.0,
    velocity=1.2,           # mg/dL/min
    insulin_on_board=1.4,   # units
    carbs=0.0,
    current_time=0.0,
)

print(decision.insulin_delivered)
print(decision.action)
print(decision.reasoning)

Next Improvements¶

Verify parameters against updated manuals and peer‑reviewed sources.
Add explicit citations into emulator output metadata.
Add regression tests against published outcome ranges.