CASIA OpenIR  > 毕业生  > 博士学位论文
基于感兴趣区域(ROI)的锥束CT重建
其他题名Cone-beam CT Reconstruction based on Region-of-Interest(ROI)
左年明
学位类型工学博士
导师蒋田仔 ; 潘晓川
2007-05-24
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业模式识别与智能系统
关键词锥束ct Ct重建 滤波反投影 局部成像 Ct(Computed Tomography) reconstruCtion Cone-beam Ct Roi Imaging Local Tomography Fbp(Filtered-backprojection) Bpf(Backprojection-filteration)
摘要自上个世纪七十年代第一台商用~CT~扫描机器面世以来,CT~成像技术已经经历了三十多年的长足发展,无论是成像的速度和精度,还是硬件对射线的利用率以及扫描轨迹的灵活性等,都得到了实质性的发展和提高。八十年代的FDK算法的提出,实现了三维意义的~CT~重建,但它是一种近似的三维重建,只有对锥角较小的扫描束才比较适用(例如4排或者8排的探测器形成锥角)。随着硬件技术的发展以及现代医疗对成像速度和精度要求的进一步提升,探测器出现了64排甚至是256排的情形,此时的射线束的锥角相对于前者已经很大了,不能简单的用三维空间的平行线束来近似模拟,因此,必须依靠真正三维意义上的精确重建算法。另一方面,现代医疗技术为了减少X-射线对正常器官组织的辐射,希望尽量能对病灶区域(感兴趣区域,ROI)集中成像,这就要求利用尽量少的扫描射线来获得ROI的。这就产生如下两个问题:首先,射束的视域(Field of View)应该尽量小,能罩住ROI,不一定要求罩住整个物体;第二,扫描的轨迹应该尽可能短,而且针对ROI选择有针对性的轨迹,不仅仅限于传统的圆弧,螺旋轨迹等。尤其是在成像和放疗一体化的~CT~机中,放疗过程中的轨迹很可能根据实际情况有多样化的需求,所以上述第二点表现得尤为重要。因此,为了解决上述的系列问题,我们需要真正意义上的三维重建算法,能针对ROI精确成像,即使投影数据有少量的截断;并且针对不同的扫描轨迹都有灵活有效的重建算法。 本论文的主要创新点表现如下:针对含有纵向截断和横向截断的投影数据,提出了基于圆柱面探测器的锥束~CT~重建算法,不但能处理投影数据的纵向截断问题,也能处理投影数据的横向截断问题,从而更好的进行ROI的重建;在较大锥角的情形下,为了减少相邻的不同射线束对同一探测单元的影响,从而提高探测精度,提出了球心在发射源的球面形探测器,并针对球面探测器设计了一种均匀的探测单元设计方法,最后,对此探测器也给出了相应的精确重建算法;提出了一般结构的直线圆弧扫描轨迹,使得能更方便的针对~ROI~重建,并对此类轨迹给出了一般的精确重建算法;对系列基于弦线的重建算法给出了系统的评估,包括重建精度以及噪声特性等属性的定性和定量的比较;为了精确定量的研究重建算法的噪声属性,我们单独的研究了算法过程中的加权,反投影和滤波等每一个步骤的噪声特性;作为最近该领域的两个重大突破,Katsevich和Zou-Pan算法,我们研究了它们之间的内在联系,并指出,在选取同一滤波平面的情形下,它们在数学表达上的一致性;针对~FOV~不能罩住~ROI~的情形(数学上已被证明不能精确重建),我们仍然利用不完整的投影数据进行了局部重建,得到了~ROI~区域物体的清晰的边界轮廓,这可为临床诊断提供有力的帮助。
其他摘要Through over 30 years development, medical CT (Computed Tomography) imaging technology has achieved substantial enhancement both in efficiency and precise,from 2-D (Two Dimensional), in the modest begin by use of which to pile up toform a 3-D (Three Dimensional) volume image, to virtually 3-D cone-beam reconstructionin recent years.In the middle of 1980's, a novel 3-D reconstruction method, FDK algorithm, has beendeveloped, which was a generalized FBP method from 2-D to 3-D and played a criticallyimportant role not only in theoretical consideration but for commercial application, although that was an approximate method. After that, many researchers were involved tostriving for virtually 3-D reconstruction, since that is a necessity for cone-beamreconstruction causing a bigger cone angle in the condition of multi-row detector.More considerably, modern medical (clinical) standard requires as less as possible radiation towardsnormal tissues surrounding the diseased region, which would reduce unnecessary harmto the healthy organs, therefore, it is in an urgent need of a specific reconstruction methodfor a special ROI (Regions of Interest), required to accommodate incomplete projectiondata with serious truncations. Such a requirement, on the other hand, is also reasonablefor the emerging imaging-radiotherapy CT machine bringing referable images as by products in the radiotherapy process.The main contributions in this PhD dissertation are as following,Presented a serial of chord-based cone-beam reconstruction algorithms, some of whichcan obtain exact ROI images from projection data with serious longitudinal and transversetruncations; Conducted a complete evaluation for the algorithms above, including the noise properties and reconstruction precise. Additionally, we have also achieved a quantitativedescription for the noise property in each reconstruction step, embracing weightingstep, backprojection step and filtering step; Introduced a spherical detector and the corresponding uniform sampling method on the detector and reconstruction algorithm,which could reduce the cross-talk absorption in the adjacent bins in the detector; Presented a general reconstruction algorithm for a generalized line-circle scanning trajectory, which is more specific for ROI imaging; Introduced a conceptual way to understand the relations between the Katsevich algorithm and Zou-Pan algorithm; Implemented chord-based local tomography, which can obtain the exact contourof ROI as a valuable reference for clinical diagnosis.
馆藏号XWLW1115
其他标识符200418014628021
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/5970
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
左年明. 基于感兴趣区域(ROI)的锥束CT重建[D]. 中国科学院自动化研究所. 中国科学院研究生院,2007.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_20041801462802(4476KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[左年明]的文章
百度学术
百度学术中相似的文章
[左年明]的文章
必应学术
必应学术中相似的文章
[左年明]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。