基于区分矩阵与近似的支撑矢量机的医学影像库分类策略

针对医学影像库信息量大、关联信息多、对象复杂等特点,将基于区分矩阵的属性约简算法与一种近似的支撑矢量机算法相结合实现了对医学影像库的正常、异常分类。基于区分矩阵的属性约简算法有效地降低了医学影像库的维度,而非线性的近似支撑矢量机算法则克服了标准支撑矢量机在实际应用中表现出来的算法速度慢、算法过于复杂而难于实现以及检测阶段运算量大等缺陷。实践证明了该方法的确具备简单、快速,高效的特点。

A strategy of medical image database classification based on discernibility matrix and PSVM

In this paper, an algorithm of attribute reduction based on discernibility matrix and a proximal support vector machine （PSVM） is integrated and used to implement a classification of medical image database. The attribute reduction algorithm is used to reduce useless and interfering attributes of medical images, and the PSVM is applied to classify the medical images as normal and abnormal class. The PSVM not only runs faster than standard support vector machine classifiers, hut also is easy to implement with satisfactory result for lower hardware. It is proved by experiments that our strategy do have some good features such as simpleness, speediness and high efficiency.