An artificial neural network using the Kohonen Layer architecture achieved QRS classification performance equivalent to sophisticated literature methods when adapted to patient-specific normal patterns.
This work shows an artificial neural network using the Kohonen Layer architecture in a modified approach to support supervised learning, and the evaluation of its performance in the classification of QRS complexes of the electrocardiogram (EGG) from patients with cardiac arrhythmias. A second aim of this study was to investigate the ability of ANN to classify QRS complexes when the original data samples are used as input variables. The classifier was developed and tested with the MIT-BIH Arrhythmia Database. The obtained results become equivalent to the most sophisticated methods in the literature when input data are properly pre-processed and the final classifier is allowed to adapt to the normal pattern of each analyzed patient.
Melo et al. (Mon,) conducted a other in Cardiac arrhythmias. Artificial neural network (Kohonen Layer architecture) vs. Sophisticated methods in the literature was evaluated on Classification of QRS complexes. An artificial neural network using the Kohonen Layer architecture achieved QRS classification performance equivalent to sophisticated literature methods when adapted to patient-specific normal patterns.