Bayesian Network Learning for Detecting Reliable Interactions of Dose-Volume Related Parameters in Radiation Pneumonitis

Due to the high fatality rate of patients with radiation pneumonitis (RP), a complication of the radiation therapy (radiotherapy), great attention has been paid to the treatment plan of individual RP patients. Therefore, not only technological advances in the development of treatment planning systems but also new prognostic models are urgently required to lessen the complication and to predict the state of patients more accurately. The Bayesian network is a useful tool for finding interactions among features and for developing prognostic models that enable physician to predict the outcome of radiotherapy. In this paper, we show the interactions among dosimetric features through Bayesian network structures and the performance of Bayesian classifiers with different search algorithms on a non-small cell lung cancer (NSCLC) dataset.