ICCESEN 2016 (1)

Performance Comparison of Artificial Neural Network Training Algorithm for Fetal Heart Rate Patterns

We attended a conference called as “3rd International Conference on Computational Science and Engineering (ICCESEN)” for the first time this year. We met with researchers who came from around the world. This conference has a special meaning for me. For the first time, I have attended a conference with my family and presented my work in English. The conference was held in Kemer, Antalya, Turkey. During the conference time, we have done both work and holiday. I would like to thank you the chairman Prof. Dr. İskender AKKURT for everything, Dr. Zehra KULUÖZTÜRK and Dr. M. Fatih KULUÖZTÜRK for their close attention.

I presented my work entitled as "Performance Comparison of Artificial Neural Network Training Algorithm for Fetal Heart Rate Patterns." The abstract of the work is given at below.

"Performance Comparison of Artificial Neural Network Training Algorithm for Fetal Heart Rate Patterns"

Abstract

In biomedical signal applications, Cardiotocography (CTG) is crucial because it contains fetal heart rate patterns which offer significant and vital signs about the fetal condition. Over the last decades, fetal heart rate patterns are classified as healthy and unhealthy. As a classifier, artificial neural network (ANN) is notable due to its powerful capabilities. For this reason, behaviors, and performances of neural network training algorithms have been investigated and compared on the classification task of the fetal heart rate patterns. Training algorithms of neural network have been categorized in five group as Gradient Descent, Resilient Backpropagation, Conjugate Gradient, Quasi-Newton, and Levenberg-Marquardt. Sensitivity, specificity, accuracy, and quality index evaluation criteria have been taken into account during performance comparison of the algorithms. According to results of this study, the best classification performance has been obtained with Conjugate gradient backpropagation with Fletcher-Reeves restarts when consumed training period of networks is not taken into account. Also, Levenberg-Marquardt backpropagation has achieved remarkable results when classification success and consumed training period of networks are considered.

Keywords:  Cardiotocography, Artificial Neural Network, Classification, Fetal Heart Rate, Pattern Recognition

The full text of the paper will be published in the first period of 2017. 

2016-10-30, Pazar