AUTOPILOT-BT: an approach towards automatic mechanical ventilation

Introduction

The clinical use of ventilators is limited due to a huge variety of different ventilation methods. The clinician – often under high cognitive load from the complicated technical equipment on an ICU – just uses a small subset of available parameter settings. The aim of the present study was to develop a closed-loop ventilation controller based on mathematical models and fuzzy logic.

Methods

The system was designed to track a desired end-tidal CO₂ pressure (PaCO₂), to find a PEEP leading to maximum estimated respiratory system compliance and to maintain the arterial oxygen saturation (SaO₂) at an optimal level. We developed a program in LabView (National Instruments, Austin, TX, USA) on a laptop that is able to read the internal data of a ventilator (Evita 4; Dräger Medical, Germany) in real time. Respiratory signals (for example, SaO₂) are acquired from monitoring. Discrete measurements (for example, PaO₂) are either assumed constant until next measurement or are interpolated using a model-based approach evaluating, for example, the etCO₂ data. The course of etCO₂ following the setting of optimal frequency was evaluated to calculate the time required for equilibration of etCO₂.

Results

A module automating the initial settings of the ventilator according to local ICU rules is realized. Modules were added that optimize breathing frequency with respect to PaCO₂/etCO₂ and FiO₂ according to SO₂ whenever no PaO₂ is available. A lung simulator (Michigan Instruments Inc., Grand Rapids, MI, USA) connected with the LS4000 (Dräger Medical) was used to evaluate the system. Exemplary results are presented in the figures, which show the minute volume/etCO₂ relationship (Figure 1) and a parametric fit of etCO₂ data (Figure 2). The adjustment of the frequency is based on the current etCO₂ model.

Figure 1

Minute volume/etCO₂ relationship.

Figure 2

Parametric fit of etCO₂ data.

Conclusion

Automation is a 'sine qua non' to achieve optimal patient individualized ventilation support. Our system is enabled to evaluate a therapeutic strategy and to base the settings of the ventilator on current trends/drifts observed in the data.