EKG feature extraction and classification for Myocardial Infarction Diagnosis

Myocardial Infarction Diagnosis (MID) is commonly based in eletrocardiography recording (EKG) and performed in healthcare institutions by skilled professionals. Nowadays, EKG based diagnosis are widespread and as a consequence increasing the workload of health professionals, leading to possible wrong diagnosis. The paper proposes an automatic classification system of EKG signals, to help health professionals in their daily tasks, aiming for a first diagnose screening tool. The work presented in this paper is twofold, i.e., in the first part EKG descriptors are extracted from the EKG signals. In the second part, the extracted descriptors are used as input to a classification system based on computational intelligence methods. These methods include fuzzy systems, neural networks and support vector machines. The EKG raw signals were pre-processed using a bank of Butterworth filters, for preparation towards the extraction of Q, R and S waves, along with the inter-beat rate. The PTB (Physikalish-Technische Bundesanstalt) Diagnostic EKG Database was used to validate the approach presented in the paper and to obtain the desired diagnose.