Knowledge Expression and Inference Based on Fuzzy Bayesian Networks to Predict Astrocytoma Malignant Degree

A modified fuzzy Bayesian network (FBN) is proposed in this study, which integrates fuzzy theory into Bayesian networks (BN) by using Gaussian mixture models (GMM) to make a fuzzy procedure. This particular procedure transforms continuous variables into discrete ones, when dealing with continuous inputs with probabilistic and uncertain nature. Based on the FBN, the fuzzy reasoning model for prediction and diagnosis can be designed. To validate our method, two models are built and used to classify the astrocytoma malignant degree, which can be modeled by probability quantitatively. The experiment results show that the model fusing both low-level image features and high-level semantics outperforms the one only using low-level image features with very promising results. This FBN model also provides knowledge expression in predicting astrocytoma malignant level. This study provides a novel objective method to quantitatively assess the astrocytoma malignancy level that can be used to assist doctors to diagnose the tumor