基于关联分类算法的医学图像数据挖掘

Medical images data mining using classification algorithm based on association rule

目的利用关联分类算法，构造医学图像分类器，对未知类型的脑肿瘤图像进行自动判别和分类，以帮助临床医生进行脑疾病的诊断和治疗。方法对医学图像经过预处理后进行特征提取，再将提取的特征离散化后放到事务数据库中作为关联分类规则的输入，然后利用改进的Apriori算法构造医学图像分类器。结果构造了医学图像分类器，用已知类型的图像训练分类器挖掘满足约束条件的关联规则，然后利用发现的关联规则对未知类型的医学图像进行分类以判断脑肿瘤的良恶性。结论利用关联分类算法可以有效地挖掘医学图像特征，进而构造图像分类器，实现脑肿瘤良恶性的自动判别。

Objective In order to assist clinicians in diagnosis and treatment of brain disease, a classifier for medical images which contains tumors inside, based on association rule data mining techniques was constructed. Methods After a pre-processing phase of the medical images, the related features from those images were extracted and discretized as the input of association rule, then the medical images classifier was constructed by improved Apriori algorithm. Results The medical images classifier was constructed. The known type of medical images was utilized to train the classifier so as to mine the association rules that satisfy the constraint conditions. Then the brain tumor in an unknown type of medical image was classified by the classifier constructed. Conclusion Classification algorithm based on association rule can be effectively used in mining image features, and constructing an image classifier to identify benign or malignant tumors.