3D Vectorcardiographic Machine Learning for Classification Cardiovascular Disease

EasyChair Preprint no. 10552

Abstract

This study aims at comparing the performance of computational intelligence methods to classify ECG data in normal (NORM), myocardial infarction (MI), and ST-T change (STTC) groups using the XYZ ECG coordinates as input. The ECGs of 7146 patients were randomly selected from the PTBXL Database to produce a balanced dataset. The multi-layer perceptron model achieved 99.99% accuracy during the training and 99.80% in testing. The convolutional neural network model achieved 96.07% accuracy during the training and 83.26% in testing. The long short-term memory (LSTM) model achieved 99.90% accuracy during the training and 89.00% during the test. Also, the LSTM model applied to 10-fold produced an average accuracy of 94.03 ± 1.83%. In conclusion, this study provides an effective framework for the automated detection of MI and STTC on ECG. Specifically, it classifies NORM, MI, and STTC with more than 94% accuracy and hence can be employed in clinical settings.

Keyphrases: Artificial Neural Networks, myocardial ischemia, Vectorcardiography

@Booklet{EasyChair:10552,
  author = {Lucenildo Cerqueira and Jurandir Nadal},
  title = {3D Vectorcardiographic Machine Learning for Classification Cardiovascular Disease},
  howpublished = {EasyChair Preprint no. 10552},

  year = {EasyChair, 2023}}