基于图像分割的肺结节CT图像哈希检索

基于相似图像的肺结节CT图像检索辅助诊断对肺结节的发现有着重要的作用。肺结节的诊断难度较大,通常需要充分利用图像的边缘、分叶、毛刺、纹理等各类信息。文中针对目前基于哈希方法的肺结节检索中存在的不能充分利用图像分割信息从而导致部分信息丢失问题做出了改进,提出了一种基于图像分割的肺结节图像哈希检索方法。实验结果表明,在72位哈希码长度时,达到了85.3%的平均准确率。并且,将文中图像分割模块应用于其他哈希检索方法时,平均准确率皆有一定的提升。

CT image hash search of lung nodules based on image segmentation

CT image retrieval-assisted diagnosis of lung nodules based on similar images plays an important role in the discovery of pulmonary nodules.The diagnosis of pulmonary nodules is difficult,and it is usually necessary to make full use of various information such as edges,lobes,burrs,and textures of the image.In order to improve hashing method,which can not make full use of image segmentation information and lead partial information loss,a method of image retrieval based on image segmentation for lung nodule image is proposed.The experimental results show that the average accuracy of 85.3%is achieved at 72-bit hash code length.Moreover,when the image segmentation module of this paper is applied to other hash retrieval methods,the average accuracy rate is improved.