Abstract:
Arrhythmia is the most common cardiovascular disease (CVD) taking a toll of approximately 10 million cases per year in India. ECG analysis is the most prominent way of detecting the Arrhythmia abnormalities. ECG being complex with many crest and trough, it would be of great help if automated analysis of ECG can done. In this work we propose a framework for prediction of arrhythmia risk by statistically analyzing the ECG of the user. We have proposed novelistic way for training and testing for multi-stage classification to improve the accuracies and sensitivities of the model. Further we also consider the time aspect of the model since the model should be able to put in use in real-time applications. We focuses on to build a model that can be combined with personal holter machines so that it can act as a precursor to the consultation with the doctor.
We have used MIT-BIH Arrhythmia Benchmark dataset following the AAMI recommendations and the work corresponds to inter-patient paradigm.
