基于CS和ICA的多能X射线医学图像目标分离算法研究

发布时间：2018-06-27 00:09

本文选题：独立成分分析 + 压缩感知　；参考：《南京邮电大学》2016年硕士论文

【摘要】：X射线医学成像可以被用于观察到患者体内病变组织,因此它对医学诊断有重要参考价值。针对传统医学X射线图像噪声强、层次感差和器官组织重叠的问题,本文提出利用独立成分分析(Independent Component Analysis,ICA)从多能X射线图像中进行降噪和重叠目标分离成像研究。压缩感知(Compressed Sensing,CS)基于信号的稀疏理论,利用随机观测矩阵以远低于Nyquist采样频率的抽样速率处理信号,可以大大缩短采样时间、减少计算量。压缩感知的这个突出的优点可以被用来提高ICA进行分离成像的性能和速度。本文研究一种新型的X射线成像方法,利用独立成分分析从多能X射线医学图像中将重叠目标分离成像,并结合压缩感知以提高算法运行速度和分离成像质量。具体研究方法为:根据被拍摄物体的物质组成,拍摄至少三幅多能X射线图像组成图像序列;应用CS选取稀疏基将多能X射线医学图像进行稀疏表示;然后利用ICA将多能X射线图像的稀疏表示进行盲源分离得到独立源;最后采用正交匹配追踪算法(Orthogonal Matching Pursuit,OMP)将独立源重构达到分离成像。研究结果表明:使用CS结合ICA与仅使用ICA进行分离成像相比较,前者比后者在算法运行时间减少46.14s(23.3%)、内存占有率降低到75%、重构图像峰值信噪比(Peak Signal to Noise Ratio,PSNR)提高2.665dB、边缘梯度提高0.001、信息熵提高0.09。利用独立成分分析可以从多能X射线图像中将重叠目标分离成像,结合压缩感知技术会使分离成像的速度和性能有较高提升。
[Abstract]:X-ray medical imaging can be used to observe the pathological changes in patients, so it has important reference value for medical diagnosis. Aiming at the problems of high noise, poor stratification and overlapping of organs and tissues in traditional medical X-ray images, an independent component analysis (ICA) is proposed to study the de-noising and overlapping target separation imaging from multifunctional X-ray images. Compressed sensing CS (CS) is based on the theory of signal sparsity, using random observation matrix to process signals at a sampling rate far below Nyquist sampling frequency, which can greatly shorten the sampling time and reduce the amount of computation. This outstanding advantage of compression sensing can be used to improve the performance and speed of ICA separation imaging. In this paper, a new method of X-ray imaging is studied, in which overlapping targets are separated from multifunctional X-ray medical images by independent component analysis (ICA), and compression perception is combined to improve the speed of the algorithm and the quality of the separation imaging. The specific research methods are as follows: according to the material composition of the object being photographed, at least three multi-energy X-ray images are taken to constitute the image sequence, and the sparse basis is selected by CS to represent the multi-energy X-ray medical image sparsely. Then the sparse representation of multi-energy X-ray image is separated by ICA to obtain the independent source, and the orthogonal matching pursuit algorithm (orthogonal matching pursuit OMP) is used to reconstruct the independent source to achieve the separation imaging. The results show that the use of CS combined with ICA is compared with the separation imaging using ICA alone. The former is 46.14s (23.3%) less than the latter, the memory occupancy is reduced to 75, the peak signal to noise PSNR (Peak signal to noise ratio) is increased by 2.665dB, the edge gradient is increased by 0.001, and the information entropy is increased by 0.09. Using independent component analysis (ICA), overlapping targets can be separated and imaged from multifunctional X-ray images. Combined with compression sensing technology, the speed and performance of separation imaging can be improved.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R814;TP391.41

【相似文献】