Semi-supervised learning of probabilistic models for ECG segmentation

We present a novel semi-supervised learning algorithm, based upon the EM algorithm for maximum likelihood estimation, which can be used to learn probabilistic models from subjectively labelled data. We demonstrate the method on the task of automated ECG segmentation, with a particular emphasis on the accurate measurement of the QT interval. In addition we discuss the use of wavelet methods for the representation of the ECG, and advanced duration modelling techniques for hidden Markov models applied to ECG segmentation.