Evaluation of a Gaussian Mixture Model forgenerating synthetic ECG signals during anangioplasty procedure

Abstract

Mathematical models have intensively been used in the generation of synthetic electrocardiogram (ECG) signals. They are used to test and optimize different algorithms for patient monitoring and heart disease detection. In this work, we present a Gaussian mixture model that allows the generation of heartbeats with ischemic alterations, induced by an occlusion procedure performed in the right coronary artery. Realizations obtained using the Gaussian mixture model were compared against real signals using cross-correlation, obtaining average values equal to 94.2%. Confidence intervals, at 99% level, calculated using ST values from the synthesized heartbeats agreed with the real ST values in 100% cases for each group (ECG Lead and ST-level). In this sense, the model proposed allows the synthesis of heartbeats with complex morphologies that cannot be obtained with more traditional models.