Validating ECG signals using a single-layer ANN

The Intelligent Anesthesia Monitor (IAM) project aims to examine and implement the steps necessary to build an anesthesia monitor which advises an anesthetist during an operation, alarms him only when necessary and presents him with useful information. One of the first problems encountered is that the measured signals (like the electrocardiogram (ECG) and blood pressure) are often corrupted with artefacts. Here, the authors present an algorithm which tries to validate the ECG, based on its shape. First, an unsupervised single-layer artificial neural network (ANN) is used to extract features from the ECG. Next, the anesthetist is asked to classify a few seconds of the signal as valid. Finally, the authors compare the features extracted from the validated signal with the features extracted from the current signal. To keep the algorithm adaptive, a mechanism is implemented to allow that new validated signals will eventually become the new reference for comparison. Preliminary results tested on the ECG indicate that this method is well-suited for this problem, but further testing has to be performed