本文關(guān)鍵詞：有限投影數(shù)據(jù)CT圖像迭代重建技術(shù)研究　出處：《南方醫(yī)科大學(xué)》2017年博士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： CT成像 稀疏角度 有限角度 全變差 非局部平均 迭代初始圖像

【摘要】：計(jì)算機(jī)斷層成像(Computed Tomography,CT)中,過高的X射線劑量危害人體。增大間隔采樣投影數(shù)據(jù)可以達(dá)到降低劑量的目的,但不再滿足經(jīng)典FBP解析精確重建條件,而迭代重建方法表現(xiàn)出它的優(yōu)勢(shì)。Sidky等人提出了 TV最小化約束ART重建圖像的算法(簡稱ART-TV),但TV算法存在的問題是,重建圖像會(huì)出現(xiàn)階梯偽影。另外,非局部平均(NLM)濾波算法去噪效果優(yōu)于TV,前人將NLM濾波融入在CT重建中,取得優(yōu)于ART-TV的重建結(jié)果。少于180°掃描范圍的有限角度成像也是一種降低劑量手段。FBP算法重建圖像含有明顯條形偽影,并不均勻分布在圖像中。迭代方法可以重建出更好質(zhì)量圖像,但仍存在不足。尚無NLM算法在有限角度CT重建成功應(yīng)用的報(bào)道。本文研究工作包括以下四個(gè)方面:第一,改進(jìn)并實(shí)現(xiàn)一種快速優(yōu)質(zhì)FBP重建算法。全角度CT重建中,考慮投影數(shù)據(jù)中零數(shù)據(jù)代表對(duì)應(yīng)的X射線只穿過空氣,因此,在對(duì)每個(gè)重建像素進(jìn)行反投影操作時(shí),依次判斷重建像素點(diǎn)對(duì)應(yīng)的探測(cè)器線積分投影數(shù)據(jù)是否為零,若是,直接對(duì)像素點(diǎn)置零,不必進(jìn)行剩余反投影操作。實(shí)驗(yàn)表明,算法速度更快,空氣像素值更準(zhǔn)。有限角度CT中,若物體軸對(duì)稱,利用FBP重建圖像中無偽影的物體邊緣確定和修補(bǔ)另一對(duì)側(cè)結(jié)構(gòu)信息,此確定對(duì)側(cè)邊緣過程用到以上所提算法,根據(jù)對(duì)稱軸找到對(duì)側(cè)輪廓并對(duì)對(duì)稱輪廓信息進(jìn)行恢復(fù)。優(yōu)化的FBP圖像作為ART-TV算法初始圖像,實(shí)驗(yàn)表明,迭代收斂速度更快。第二,改進(jìn)并實(shí)現(xiàn)一種ART-NLM算法用于有限角度CT重建。根據(jù)局部偽影在圖像中的位置,判斷重建像素位于哪一區(qū)域,若位于偽影區(qū)域,則利用其對(duì)稱象限內(nèi)的像素進(jìn)行NLM濾波,否則利用重建像素周圍的像素進(jìn)行NLM濾波,有效恢復(fù)偽影區(qū)域內(nèi)的正常信息。如果物體對(duì)稱關(guān)系不好,則在后續(xù)迭代重建中改用ART-TV有效避免偽像產(chǎn)生。實(shí)驗(yàn)結(jié)果表明,提出的ART-NLM/TV在偽影消除方面效果更佳。第三,改進(jìn)并實(shí)現(xiàn)一種ART-ATpV算法用于稀疏角度CT重建。提出基于自適應(yīng)TpV的迭代重建算法,改進(jìn)算法為梯度圖像的p(0≤p≤l)范數(shù),根據(jù)每個(gè)重建像素的屬性確定自身p值。當(dāng)像素點(diǎn)位于均勻區(qū)域時(shí),p接近1,有效去除噪聲,而當(dāng)像素點(diǎn)位于邊緣區(qū)域,p接近0,有效保護(hù)邊緣結(jié)構(gòu)信息。實(shí)驗(yàn)結(jié)果表明,新方法可以有效去除噪聲的同時(shí)最大程度保護(hù)邊緣結(jié)構(gòu)信息。第四,改進(jìn)并實(shí)現(xiàn)一種ART-TV/SL0算法用于稀疏角度CT圖像迭代重建。提出基于TV結(jié)合平滑L0范數(shù)(Smooth L0,SL0)的迭代重建算法,即,ART重建和非負(fù)約束后,進(jìn)行TV最小化優(yōu)化圖像,之后采用梯度圖像的SL0最小化進(jìn)一步去除噪聲和偽影。實(shí)驗(yàn)結(jié)果證明,改進(jìn)算法平衡了噪聲和偽影去除以及結(jié)構(gòu)邊緣信息保護(hù)。
[Abstract]:In computed Tomography (CTT), too high X-ray dose is harmful to human body. Increasing the interval sampling projection data can achieve the purpose of reducing the dose. But it no longer satisfies the classical FBP analytical exact reconstruction condition. The iterative reconstruction method shows its advantages. Sidky et al. put forward a TV minimization constrained ART image reconstruction algorithm (referred to as ART-TVT), but the problem of TV algorithm is. Step artifacts appear in reconstructed images. In addition, the non-local average NLM filter algorithm is better than TV-based filtering algorithm in denoising effect. Previous researchers incorporated NLM filter into CT reconstruction. The finite angle imaging with less than 180 擄scanning range is also a method of reducing dose. The iterative method can reconstruct the image with better quality. However, there are still some shortcomings. There is no report on the successful application of NLM algorithm in finite angle CT reconstruction. The research work in this paper includes the following four aspects: first. Improve and implement a fast and high quality FBP reconstruction algorithm. In full angle CT reconstruction, the zero data in projection data represent the corresponding X-ray only through the air, so. When each reconstructed pixel is backprojected, the detector line integral projection data corresponding to the reconstructed pixel point is judged to be zero or not, and if so, the pixel point is directly set to zero. The experiments show that the algorithm is faster and the air pixel value is more accurate. In finite angle CT, if the object is axisymmetric. Using FBP to reconstruct the edge of the object without artifacts in the image to determine and repair another pair of side structure information, this determination of the opposite edge of the process using the above algorithm. According to the symmetry axis to find the opposite contour and restore the symmetrical contour information. The optimized FBP image as the initial image of the ART-TV algorithm, the experimental results show that the iterative convergence speed is faster. Second. A ART-NLM algorithm is improved and implemented for finite angle CT reconstruction. According to the location of local artifacts in the image, the reconstruction pixels are located in which region, if located in the artifact region. Then the pixels in the symmetric quadrant are used for NLM filtering, otherwise, the pixels around the reconstructed pixels are used for NLM filtering, which can effectively restore the normal information in the artifact region, if the symmetry relation of the object is not good. The experimental results show that the proposed ART-NLM/TV is more effective in artifact elimination. An improved and implemented ART-ATpV algorithm for sparse angle CT reconstruction is proposed. An iterative reconstruction algorithm based on adaptive TpV is proposed. The improved algorithm is the p0 鈮，

本文編號(hào)：1356454