基于粗糙集特征级融合的肺结节检测算法

以肺结节的检测为研究目标,针对肺结节特征级融合检测算法中存在特征结构不合理和特征表达不紧致两个问题,提出了一种基于粗糙集特征级融合的肺结节检测算法,该算法首先分析肺部CT影像的医学征象,提出了六个新的三维特征,并综合其他二维和三维特征共42维特征分量共同量化ROI;然后基于粗糙集对提取的特征集合进行5次特征级融合实验;最后利用网格寻优算法优化核函数的SVM作为分类器进行肺结节识别.以70例肺结节患者的肺部CT影像为原始数据,通过4组对比实验验证算法的有效性和稳定性,实验结果表明,经过粗糙集特征级融合的肺结节检测算法识别肺结节的能力得到了有效提升.

Lung Nodules Detection Based on Rough Sets at Feature Level

Based on the detection of lung nodules as the research target, in terms of the two problems of lung nodules detection algorithm at the feature level fusion that features structure is not reasonable and feature expression is not tight, this paper proposes a lung nodule detection algorithm based on rough sets at feature level fusion, the algorithm firstly analyze medical signs of lung CT image, six new 3d characteristics was proposed, and other 2d and 3d features as a total of 42 features quantitative ROI; then based on rough sets to extract the feature sets for 5 times experiments; at last the grid optimization algorithm was used to optimize the kernel function of SVM as classifier to identify the lung nodules. Based on 70 cases of lung nodules in patients with lung CT images as the original data, through four groups of experiments verify the validity and stability of the algorithm, and the experimental results show that after the lung nodules detection algorithm of rough set feature level fusion effectively improved the recognition ability of lung nodules.