基于贝叶斯决策树的医学图像分类方法研究

AN APPROACH OF MEDICAL IMAGE CLASSIFICATION BASED ON BAYESIAN DECISION TREE

随着PACS系统在我国各医院的普及,PACS数据库中存储了大量的医学图像信息,如何把这些图像进行分类来提供相似病例图片,从而为临床诊断提供辅助帮助已成为研究热点.关于医学图像的分类,已有很多学者从不同方面用不同方法进行了研究.本文使用贝叶斯决策树的方法对PACS数据库进行教据挖掘,实现医学图片的分类.贝叶斯决策树不仅能够提高分类的准确率,而且能够处理不一致,不完整数据等"脏数据",本文充分发挥了贝叶斯方法和决策树方法的优点,通过对肺癌图片进行良性、恶性分类,证明了本方法的有效性.

Along with the PACS system used in many hospitals, the PACS database saved the massive medical picture information. How to carry classification on these pictures to provide the similar image of illness and provide the assistance help for the clinical diagnosis have become a new research field. About the medical image classification, many scholars have conducted the research from the different aspects with the different method. In this paper, we using the method which based on Bayesian -decision tree, we mine the data for the PACS database and classify the medical picture. Bayesian -decision tree not only can enhance the classification accurately, but also can process inconsistent and the incomplete data which called the ＂dirty date＂. This method has displayed the advantage of the Bayesian method and the Decision tree fully. In this article, we has proven this method validity by classified the benign and the malignant of the lung cancer picture.