A neural network model for diagnosis of critical patients receiving mechanical ventilation

Artificial Neural Networks (ANNs) have become powerful tools for medical decision making. While they are able to extract decision strategies from input data, they can improve their own performance from the specific patient context. In this paper the authors present an ANN based on the PATRICIA system research work. This system has been built to obtain five categories of diagnoses for ICU patient receiving mechanical ventilation. The ANN model has been trained from 30 data spread sheets including information about ventilation, oxygenation, acid-base balance and other clinical parameters. The authors' next step will be to improve the capabilities of the system by using 60 patients under continuous monitoring.