Three Classification Tools Help Diagnose Heart Problems

To do so, the researchers sourced data from heart disease databases on the Physionet website, a site dedicated to medical data of various diseases and their study. They then processed the signals using three different approaches: Random Forests, Logistic Model Tree (LMT), and multilayer perceptron neural network - to validate the diagnostic conclusions.

The researchers used both linear (time domain and frequency domain) and nonlinear measures of heart rate variability for accurate classification of certain cardiac diseases. The classification results indicated that the combination of linear and nonlinear measures is a better indicator of heart diseases than linear or nonlinear measures alone, and the researchers were ultimately able to obtain a heart disease classification accuracy of 98.17%. The study was published in the July 2010 issue of the International Journal of Electronic Healthcare (IJEH).

"Heart rate and Heart Rate Variability (HRV) are important measures that reflect the state of the cardiovascular system. HRV analysis has gained prominence in the field of cardiology for detecting cardiac abnormalities,” said lead author Kumar Vimal, Ph.D., of the department of biomedical engineering. "Short-term variability in heart rate might be looked at low and high frequency electrical changes; indeed, the low frequency/high frequency ratio has been found to be the most influential HRV determinant of death and could help to identify patients at risk.”

Random forest is an ensemble classifier that consists of many decision trees and outputs the class that represents the mode by individual trees; a LMT is an algorithm for supervised learning tasks which is combined with linear logistic regression and tree induction; and a multilayer perceptron is a feedforward artificial neural network model that maps sets of input data onto a set of appropriate output.

Related Links:
PSG College of Technology
Physionet