基于加权复杂网络聚类的医学图像分类器研究

Research of medical image classify based on weighted complex network cluster

为了建立高效的肿瘤自动诊断系统,克服因医学MIR图像的复杂性带来的直接从图像中看出肿瘤及良、恶性质的困难,结合复杂网络社团划分的部分理论成果和K-mean聚类算法的思想,提出了基于加权复杂网络聚类的医学图像分类器。该分类器对医学图像进行预处理,建立图片特征库,构建图片加权复杂网络,在此基础上根据网络节点的加权网络特征值和连接度选取初始聚类中心进行聚类,有效地克服了传统K-mean聚类算法对初始化选值敏感性的问题,从而大大提高了分类精度。实验通过对某医院PACS系统中的部分MIR脑部图片进行分类,表明了该方法的分类精度比传统的K-mean聚类算法平均提高了8%左右。

Detecting tumor in medical MIR images is a difficult task because of complexity in the image. This brings the necessity of creating automatic tools to find whether a image present tumor or not. This proposed medical image classify based on weighted complex network cluster after analyzing advantages and disadvantages of the traditional K-mean clustering algorithm and the new theory results achieved in the field of complex networks. The experimental results show that this classify can find clustering centers better based on the weighted complex networks feature of nodes and it is robust to initialization, so the quality of clustering is improved 8% than the traditional K-mean clustering algorithm.