Prediction of periventricular leukomalacia.
Part II: Selection of hemodynamic features
using computational intelligence
Biswanath Samanta
a,
*
, Geoffrey L. Bird
b
, Marijn Kuijpers
g
,
Robert A. Zimmerman
d
, Gail P. Jarvik
f
, Gil Wernovsky
b
,
Robert R. Clancy
e
, Daniel J. Licht
e
, J. William Gaynor
c
,
Chandrasekhar Nataraj
a,
*
a
Department of Mechanical Engineering, Villanova University, 800 Lancaster Avenue, Villanova,
PA 19085, USA
b
Division of Critical Care Medicine and Cardiology, Children’s Hospital of Philadelphia,
Philadelphia, PA 19104, USA
c
Division of Cardiothoracic Surgery, Children’s Hospital of Philadelphia, Philadelphia, PA 19104, USA
d
Division of Neuroradiology, Children’s Hospital of Philadelphia, Philadelphia, PA 19104, USA
e
Division of Neurology, Children’s Hospital of Philadelphia, Philadelphia, PA 19104, USA
f
Department of Medicine (Medical Genetics), University of Washington, Seattle, WA 98195, USA
g
Academic Medical Center, University of Amsterdam, Department of Anesthesiology,
Amsterdam, Netherlands
Received 27 May 2008; received in revised form 8 August 2008; accepted 1 December 2008
Artificial Intelligence in Medicine (2009) 46, 217—231
http://www.intl.elsevierhealth.com/journals/aiim
KEYWORDS
Congenital heart
disease;
Computational
intelligence;
Data mining;
Decision tree;
Genetic algorithms;
Neural networks;
Periventricular
leukomalacia
Summary
Objective: The objective of Part II is to analyze the dataset of extracted hemody-
namic features (Case 3 of Part I) through computational intelligence (CI) techniques
for identification of potential prognostic factors for periventricular leukomalacia
(PVL) occurrence in neonates with congenital heart disease.
Methods: The extracted features (Case 3 dataset of Part I) were used as inputs to CI
based classifiers, namely, multi-layer perceptron (MLP) and probabilistic neural
network (PNN) in combination with genetic algorithms (GA) for selection of the most
suitable features predicting the occurrence of PVL. The selected features were next
used as inputs to a decision tree (DT) algorithm for generating easily interpretable
rules of PVL prediction.
Results: Prediction performance for two CI based classifiers, MLP and PNN coupled
with GA are presented for different number of selected features. The best prediction
* Corresponding authors. Tel.: +1 610 519 4994; fax: +1 610 519 7312.
E-mail addresses: biswanath.samanta@villanova.edu (B. Samanta), c.nataraj@villanova.edu (C. Nataraj).
0933-3657/$ — see front matter # 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.artmed.2008.12.004