Limits...
Heart rate variability dynamics for the prognosis of cardiovascular risk.

Ramirez-Villegas JF, Lam-Espinosa E, Ramirez-Moreno DF, Calvo-Echeverry PC, Agredo-Rodriguez W - PLoS ONE (2011)

Bottom Line: The results of the statistical procedure provided input to multi-layer perceptron (MLP) neural networks, radial basis function (RBF) neural networks and support vector machines (SVM) for data classification.These schemes showed high performances with both training and test sets and many combinations of features (with a maximum accuracy of 96.67%).Additionally, there was a strong consideration for breathing frequency as a relevant feature in the HRV analysis.

View Article: PubMed Central - PubMed

Affiliation: Computational Neuroscience, Department of Physics, Universidad Autonoma de Occidente, Cali, Colombia. juanfelipe.rv@gmail.com

ABSTRACT
Statistical, spectral, multi-resolution and non-linear methods were applied to heart rate variability (HRV) series linked with classification schemes for the prognosis of cardiovascular risk. A total of 90 HRV records were analyzed: 45 from healthy subjects and 45 from cardiovascular risk patients. A total of 52 features from all the analysis methods were evaluated using standard two-sample Kolmogorov-Smirnov test (KS-test). The results of the statistical procedure provided input to multi-layer perceptron (MLP) neural networks, radial basis function (RBF) neural networks and support vector machines (SVM) for data classification. These schemes showed high performances with both training and test sets and many combinations of features (with a maximum accuracy of 96.67%). Additionally, there was a strong consideration for breathing frequency as a relevant feature in the HRV analysis.

Show MeSH
Cumulative distribution plots comparison between a given feature considered in this work and the standard normal distribution.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3046173&req=5

pone-0017060-g001: Cumulative distribution plots comparison between a given feature considered in this work and the standard normal distribution.

Mentions: Up to this moment, several statistical methods have been related to feature selection for training and testing classifiers and for improving their overall performance. As most of computational methods in pattern recognition require a feature selection step, the nature of such procedure depends largely on the structure of the data. As a result, many computational approaches use parametric statistics, e.g., the so-called multivariate analysis of variance (MANOVA), which often reports adequate results for such purposes. However, we cannot assume that all the data has resemblance to a standard normal distribution at high statistical significance. Therefore, we implemented two methods in order to test the normality of the data: First, we conducted the Pearson's chi-square test; and second, the one-sample KS-test. According to our results, the data (for all the features taken into account), has a remarkably different distribution when compared to the normal distribution (with ). As both methods reported quite similar results, we concluded that non-parametric (distribution-free) statistical methods needed to be implemented at this stage. These methods, unlike parametric statistics, make no assumptions about the probability distributions of the variables being assessed. Figure 1 depicts the empirical cumulative distribution plot for a given feature (note the significant difference to the standard normal distribution).


Heart rate variability dynamics for the prognosis of cardiovascular risk.

Ramirez-Villegas JF, Lam-Espinosa E, Ramirez-Moreno DF, Calvo-Echeverry PC, Agredo-Rodriguez W - PLoS ONE (2011)

Cumulative distribution plots comparison between a given feature considered in this work and the standard normal distribution.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3046173&req=5

pone-0017060-g001: Cumulative distribution plots comparison between a given feature considered in this work and the standard normal distribution.
Mentions: Up to this moment, several statistical methods have been related to feature selection for training and testing classifiers and for improving their overall performance. As most of computational methods in pattern recognition require a feature selection step, the nature of such procedure depends largely on the structure of the data. As a result, many computational approaches use parametric statistics, e.g., the so-called multivariate analysis of variance (MANOVA), which often reports adequate results for such purposes. However, we cannot assume that all the data has resemblance to a standard normal distribution at high statistical significance. Therefore, we implemented two methods in order to test the normality of the data: First, we conducted the Pearson's chi-square test; and second, the one-sample KS-test. According to our results, the data (for all the features taken into account), has a remarkably different distribution when compared to the normal distribution (with ). As both methods reported quite similar results, we concluded that non-parametric (distribution-free) statistical methods needed to be implemented at this stage. These methods, unlike parametric statistics, make no assumptions about the probability distributions of the variables being assessed. Figure 1 depicts the empirical cumulative distribution plot for a given feature (note the significant difference to the standard normal distribution).

Bottom Line: The results of the statistical procedure provided input to multi-layer perceptron (MLP) neural networks, radial basis function (RBF) neural networks and support vector machines (SVM) for data classification.These schemes showed high performances with both training and test sets and many combinations of features (with a maximum accuracy of 96.67%).Additionally, there was a strong consideration for breathing frequency as a relevant feature in the HRV analysis.

View Article: PubMed Central - PubMed

Affiliation: Computational Neuroscience, Department of Physics, Universidad Autonoma de Occidente, Cali, Colombia. juanfelipe.rv@gmail.com

ABSTRACT
Statistical, spectral, multi-resolution and non-linear methods were applied to heart rate variability (HRV) series linked with classification schemes for the prognosis of cardiovascular risk. A total of 90 HRV records were analyzed: 45 from healthy subjects and 45 from cardiovascular risk patients. A total of 52 features from all the analysis methods were evaluated using standard two-sample Kolmogorov-Smirnov test (KS-test). The results of the statistical procedure provided input to multi-layer perceptron (MLP) neural networks, radial basis function (RBF) neural networks and support vector machines (SVM) for data classification. These schemes showed high performances with both training and test sets and many combinations of features (with a maximum accuracy of 96.67%). Additionally, there was a strong consideration for breathing frequency as a relevant feature in the HRV analysis.

Show MeSH