Artificial Metaplasticity MLP Results on MIT-BIH Cardiac Arrhythmias Data Base

This paper tests a novel improvement in neural network training by implementing Metaplasticity Multilayer Perceptron for cardiac arrhythmias classification. The proposed training algorithm is inspired by the biological metaplasticity property of neurons.The plasticity property of synaptic connections in the brain is modeled in many Artificial Neural Networks as a change in the connection weights of the artificial neurons.Artificial Metaplasticity bases its efficiency in giving more relevance to the less frequent patterns and subtracting relevance to the more frequent ones.We have applied artificial metaplasticity multilayer perceptron (AMMLP) to cardiac arrhythmias classification. The MIT-BIH Arrhythmia Database was used to train and test AMMLPs.The performance of this algorithm is tested using classification accuracy, sensitivity and specificity analysis, and ROC results.The best result obtained so far with the AMMLP algorithm is 98.25% of accuracy. A very promising result compared to the Backpropagation Algorithm (BPA) and recent classification techniques applied to the same database.