Logistic Regression and Artificial Neural Networks for Classification of Ovarian Tumors

Ovarian masses are a common problem in gynaecology. A reliable test for preoperative discrimination between benign and malignant ovarian tumors is of considerable help for clinicians in choosing approp riate treatments for patients. This study was carried out to generate and evaluate both logistic regression models and artificial neural network (ANN) models to predict m alignancy of ovarian tumors, using patient data collected at the University Hosp itals of Leuven between 1994 and 1997. The first part of the report details the stat istical analysis of the ovarian tumor dataset, including explorative univariate and multi variate analysis, and the development of the logistic regression models. The input variable selection was conducted via logistic regression as well. In the s econd part of the report, we describe the development of several types of feed f orward neural networks such as multi-layer perceptrons (MLPs) and generalized regr ession networks (GRNNs). The issue of model validation is also addressed. Our ad apted strategy for model evaluation is to perform Receiver Operating Charate ristic (ROC) curve analysis, using both a temporal holdout cross validation (CV) and multiple runs of K-fold CV. The experiments confirm that neural network classifiers have the potential to give a more reliable prediction of the malignancy o f ovarian tumors based on patient data.