OC157: Optimizing variable selection and cost using a genetic algorithm for modeling adnexal masses with Bayesian networks
Olivier Gevaert, Bart De Moor, Dirk Timmerman · Ultrasound in Obstetrics and Gynecology · 2007
Statistical methods have already proven their usefulness in diagnosis and prognosis of diseases. When the number of variables becomes large or the relationships between the variables become complex, there is a need for advanced methods to build models that help clinicians in making reliable predictions on the outcome. Possible statistical or mathematical methods include logistic regression, Support Vector Machines and Bayesian networks, which provide a method for an unbiased analysis of clinical data. We have used Bayesian networks to predict malignancy in adnexal masses. Bayesian networks offer an alternative way of modeling clinical data since a Bayesian network can explain its reasoning. We used Bayesian networks in combination with a genetic algorithm that can remove or add variables in the model such that variable selection was performed. Moreover, since each variable had a cost associated with it, we concurrently minimized the cost of the selected variables. The cost of each variable was specified by an experienced gynecologist and reflected a combination of the subjectivity, the financial cost and the time cost necessary to measure a specific variable. This method was applied to the data resulting from the International Ovarian Tumor Analysis (IOTA) multicenter study on the preoperative characterization of ovarian tumors. The first phase of IOTA resulted in a dataset consisting of 1346 masses and 1152 patients with 68 variables. We used this dataset to construct a Bayesian network and to predict the malignancy of ovarian masses while optimizing variable selection and cost. The results showed that the performance was similar to using all variables, which means that a subset can be chosen with less ‘costly’ variables without losing prediction accuracy. The models developed will be prospectively tested on new patients who are being collected in the second phase of IOTA.