An advisory system for artificial ventilation of the newborn utilizing a neural network

Abstract

A neural network has been developed to manage ventilated neonates. The network inputs are the current ventilator settings (inspiratory and expiratory times, peak inspiratory and positive end-expiratory pressures and inspired oxygen concentration), partial pressures of arterial blood gases and pH. Two hidden layers comprising 50 nodes each are employed in the network, which utilizes a standard back-propagation algorithm. The network provides the new ventilator settings as five outputs that represent the most appropriate ventilator settings projected to maintain blood gases within an acceptable range. The network has been trained using a data set derived from a rule-based expert system developed for the same purpose. Performances of both systems have been compared. The neural network is capable of learning and adapting to the individual patient's response, which in principle offers significant advantages over the rule-based system.

Predicting Extubation Outcome in Preterm Newborns: A Comparison of Neural Networks with Clinical Expertise and Statistical Modeling

“…He cautioned that because ANNs may occasionally underperform statistical models, a comparison is necessary for each new proposed application. Logistic regression has long been viewed as the "gold standard" for clinical prediction problems (3,22,23). This method has the advantage of providing coefficient estimates that can be easily interpreted, with the coefficients of the MLR model related to the mathematical odds of failing extubation.…”

“…In partic-ular, a significant number of clinical applications using neural networks have been developed (3)(4)(5)(6)(7). This growing interest can be explained by the ability of neural networks to learn directly from data, thus accommodating the intrinsic nonlinear nature of biomedical data to the detriment of rule-based systems.…”

Assessment of respiratory system mechanics by artificial neural networks: an exploratory study

“…Wilks and English (29) used ANNs, in an exploratory experiment, to classify the efficiency of respiratory patterns to predict harmful changes of the O 2 saturation in infants. Snowden et al (27) fed an ANN with blood-gas parameters and the ventilator settings that determined them to obtain advice for new ventilator settings. Orr, Westenskow, and colleagues (21,28) studied the use of intelligent alarms based on ANNs for anesthesia breathing circuits.…”

Power of breathing determined noninvasively with use of an artificial neural network in patients with respiratory failure*

“…ANNs have been applied during mechanical ventilation for breathing-pattern recognition during spontaneous breathing and PSV (26); in analysis of pressure and flow waveforms to differentiate normal from injured lungs (27); to determine ventilator settings for neonates (28); to identify respiratory abnormalities by using a pressure monitor to classify breathing patterns as effective or not and to predict changes in arterial oxygen saturation (29); to detect pulmonary embolism (30); and for assessing respiratory system mechanics during ventilatory support (31,32). We contend that another use for computer-based ANNs may be to provide the analytical basis for assessing POB.…”