Am. J. Respir. Crit. Care Med.,
Volume 160, Number 2, August 1999, 550-556
Automatic Control of Pressure Support Mechanical
Ventilation Using Fuzzy Logic
TADASHI
NEMOTO,
GEORGE E.
HATZAKIS,
C. WILLIAM
THORPE,
RONALD
OLIVENSTEIN,
SANDRA
DIAL,
and
JASON H. T.
BATES
Meakins-Christie Laboratories, Department of Biomedical Engineering, and Montreal Chest Institute,
McGill University, Montreal, Quebec, Canada
There is currently no universally accepted approach to weaning patients from mechanical ventilation,
but there is clearly a feeling within the medical community that it may be possible to formulate the
weaning process algorithmically in some manner. Fuzzy logic seems suited this task because of the
way it so naturally represents the subjective human notions employed in much of medical decision-making. The purpose of the present study was to develop a fuzzy logic algorithm for controlling pressure support ventilation in patients in the intensive care unit, utilizing measurements of heart rate,
tidal volume, breathing frequency, and arterial oxygen saturation. In this report we describe the
fuzzy logic algorithm, and demonstrate its use retrospectively in 13 patients with severe chronic obstructive pulmonary disease, by comparing the decisions made by the algorithm with what actually
transpired. The fuzzy logic recommendations agreed with the status quo to within 2 cm H2O an average of 76% of the time, and to within 4 cm H2O an average of 88% of the time (although in most of
these instances no medical decisions were taken as to whether or not to change the level of ventilatory support). We also compared the predictions of our algorithm with those cases in which changes in pressure support level were actually made by an attending physician, and found that the physicians tended to reduce the support level somewhat more aggressively than the algorithm did. We
conclude that our fuzzy algorithm has the potential to control the level of pressure support ventilation from ongoing measurements of a patient's vital signs.
