In order to deal with the complexity and uncertainty in medical diagnosis

this paper proposed a new method based on fuzzy Bayesian networks

and applied it to the classification of astrocytoma malignant degrees.Introducing a fuzzy procedure with conditional Gaussian models

aiming to transform a continuous variable into a discrete one

probability model of the classification of the degree of malignancy was established

through machine learning and with expert knowledge and sample data.Two modeling experiments demonstrate the validity of this method.Thereinto

the model fusing low-level image features and high-level semantics outperforms the one with only low-level image features.Using 60 test samples

an accuracy of 81.67% is achieved.This model provides a more objective quantitative reference and novel assistant ways for doctors to predict astrocytoma malignant level.