ECG Analysis for Automated Diagnosis of Subclasses of Supraventricular Arrhythmia

A Markov-model based technique is proposed to automatically detect a patient's disease state into different subclasses of Supraventricular arrhythmia based upon automated ECG analysis. Separate Markov-models have been developed for each subclass using a finer resolution of P-waves that takes into account left and right atria and multiple slopes of the P-waves. ECG data from Physionet database has been used to train the Markov-models. Patient’s ECG has been transformed to a probabilistictransition-graph. Graph based comparison has been used to match probabilistic-transition-graph derived from Patient’s ECG and Markov-models of the corresponding subclasses to identify the patient’s disease state in real time. The result correlates well with the physician’s diagnosis of supraventricular arrhythmia. Algorithms and sensitivity analysis have been presented.