Classification of high resolution ECG from chagasic patients with wavelet based Bayesian models

The problem of classifying multiple signals has been studied by several authors from different perspectives. Techniques such as late potential (LP) measurements in the high-resolution electrocardiogram (HRECG), abnormal intra-QRS potentials (AIQP) and time-frequency measurements have been used for evaluating signals of patients with chagasic myocarditis. The goal of these indexes is to identify different stages of the disease. However, their predictive performance is not fully satisfactory. In this work, we extend the previous developments using a wavelet-based Bayesian modelling approach for analysing and classifying unfiltered HRECG signals. We evaluate the predictive capabilities of the proposed model through a Bayesian classification based on maximum posterior probability. The technique provides considerably higher specificity and accuracy rates and may improve the predictive values of HRECG in assessing chagasic myocarditis.