中文摘要：

本研究提出一种新的融合影像低层视觉特征和语义的模糊贝叶斯网络模型。使用了高斯混合模型（GMM）对连续的视觉特征模糊化处理，解决了传统贝叶斯网络小能操作连续输入的问题，更合理地表达了具有模糊性、不确定性的专业领域的结构性知识。为了验证它的有效性，将它应用于星形细胞瘤恶性程度的分级。建立了一个概率模型。实验结果得出83．33％的正确识别率。该模型为星形细胞瘤恶性程度预测提供了新的定量而客观的辅助手段。

英文摘要：

This study proposes a form of fuzzy Bayesian networks fusing continuous low-level image features and high-level semantics, which uses Gaussian mixture models （GMM） to make a fuzzy procedure. This particular procedure will transform continuous variables into discrete ones,when dealing with continuous inputs with probabilistic and uncertain nature,so that it can settle continuous inputs that discrete Bayesian networks can＇t handle. Moreover,it describes structure knowledge in fuzzy and uncertain domain more reasonably. To demonstrate the validity of this method,we applied it to classification of astrocytoma malignant degree, and built a probabilistic model to predict astrocytoma malignant level. An accuracy of 83.33% is achieved out of testing 60 samples （30 benign and 30 malignant astrocytoma）. It provides a novel objective method to quantitatively assess the astrocytoma malignant level that can be used to assist doctors to diagnose the tumor.