光流法在放射治疗图像配准中的研究

光流法是一种可变形图像配准方法,已用来配准CT图像、四维CT（4D-CT）图像、锥形束CT（CBCT）图像、磁共振成像（MRI）图像、MRI／CT图像、PET／CT图像、SPECT图像。文章描述了光流法在放射治疗图像配准中的应用。这种技术可以使放射治疗更加有效地杀死癌细胞,同时使正常组织毒性反应更少,提高肿瘤靶区勾画的精确性,进而使肿瘤放射治疗更精确。     

A GPU-based symmetric non-rigid image registration method in human lung

Quantitative computed tomography (QCT) of the lungs plays an increasing role in identifying sub-phenotypes of pathologies previously lumped into broad categories such as chronic obstructive pulmonary disease and asthma. Methods for image matching and linking multiple lung volumes have proven useful in linking structure to function and in the identification of regional longitudinal changes. Here, we seek to improve the accuracy of image matching via the use of a symmetric multi-level non-rigid registration employing an inverse consistent (IC) transformation whereby images are registered both in the forward and reverse directions. To develop the symmetric method, two similarity measures, the sum of squared intensity difference (SSD) and the sum of squared tissue volume difference (SSTVD), were used. The method is based on a novel generic mathematical framework to include forward and backward transformations, simultaneously, eliminating the need to compute the inverse transformation. Two implementations were used to assess the proposed method: a two-dimensional (2-D) implementation using synthetic examples with SSD, and a multi-core CPU and graphics processing unit (GPU) implementation with SSTVD for three-dimensional (3-D) human lung datasets (six normal adults studied at total lung capacity (TLC) and functional residual capacity (FRC)). Success was evaluated in terms of the IC transformation consistency serving to link TLC to FRC. 2-D registration on synthetic images, using both symmetric and non-symmetric SSD methods, and comparison of displacement fields showed that the symmetric method gave a symmetrical grid shape and reduced IC errors, with the mean values of IC errors decreased by 37%. Results for both symmetric and non-symmetric transformations of human datasets showed that the symmetric method gave better results for IC errors in all cases, with mean values of IC errors for the symmetric method lower than the non-symmetric methods using both SSD and SSTVD. The GPU version demonstrated an average of 43 times speedup and ~5.2 times speedup over the single-threaded and 12-threaded CPU versions, respectively. Run times with the GPU were as fast as 2 min. The symmetric method improved the inverse consistency, aiding the use of image registration in the QCT-based evaluation of the lung.     

Reducing uncertainties in volumetric image based deformable organ registration

Applying volumetric image feedback in radiotherapy requires image based deformable organ registration. The foundation of this registration is the ability of tracking subvolume displacement in organs of interest. Subvolume displacement can be calculated by applying biomechanics model and the finite element method to human organs manifested on the multiple volumetric images. The calculation accuracy, however, is highly dependent on the determination of the corresponding organ boundary points. Lacking sufficient information for such determination, uncertainties are inevitable-thus diminishing the registration accuracy. In this paper, a method of consuming energy minimization was developed to reduce these uncertainties. Starting from an initial selection of organ boundary point correspondence on volumetric image sets, the subvolume displacement and stress distribution of the whole organ are calculated and the consumed energy due to the subvolume displacements is computed accordingly. The corresponding positions of the initially selected boundary points are then iteratively optimized to minimize the consuming energy under geometry and stress constraints. In this study, a rectal wall delineated from patient CT image was artificially deformed using a computer simulation and utilized to test the optimization. Subvolume displacements calculated based on the optimized boundary point correspondence were compared to the true displacements, and the calculation accuracy was thereby evaluated. Results demonstrate that a significant improvement on the accuracy of the deformable organ registration can be achieved by applying the consuming energy minimization in the organ deformation calculation.     

GPU-based streaming architectures for fast cone-beam CT image reconstruction and demons deformable registration

This paper shows how to significantly accelerate cone-beam CT reconstruction and 3D deformable image registration using the stream-processing model. We describe data-parallel designs for the Feldkamp, Davis and Kress (FDK) reconstruction algorithm, and the demons deformable registration algorithm, suitable for use on a commodity graphics processing unit. The streaming versions of these algorithms are implemented using the Brook programming environment and executed on an NVidia 8800 GPU. Performance results using CT data of a preserved swine lung indicate that the GPU-based implementations of the FDK and demons algorithms achieve a substantial speedup--up to 80 times for FDK and 70 times for demons when compared to an optimized reference implementation on a 2.8 GHz Intel processor. In addition, the accuracy of the GPU-based implementations was found to be excellent. Compared with CPU-based implementations, the RMS differences were less than 0.1 Hounsfield unit for reconstruction and less than 0.1 mm for deformable registration.     

A finite element method to correct deformable image registration errors in low-contrast regions

Image-guided adaptive radiotherapy requires deformable image registration to map radiation dose back and forth between images. The purpose of this study is to develop a novel method to improve the accuracy of an intensity-based image registration algorithm in low-contrast regions. A computational framework has been developed in this study to improve the quality of the 'demons' registration. For each voxel in the registration's target image, the standard deviation of image intensity in a neighborhood of this voxel was calculated. A mask for high-contrast regions was generated based on their standard deviations. In the masked regions, a tetrahedral mesh was refined recursively so that a sufficient number of tetrahedral nodes in these regions can be selected as driving nodes. An elastic system driven by the displacements of the selected nodes was formulated using a finite element method (FEM) and implemented on the refined mesh. The displacements of these driving nodes were generated with the 'demons' algorithm. The solution of the system was derived using a conjugated gradient method, and interpolated to generate a displacement vector field for the registered images. The FEM correction method was compared with the 'demons' algorithm on the computed tomography (CT) images of lung and prostate patients. The performance of the FEM correction relating to the 'demons' registration was analyzed based on the physical property of their deformation maps, and quantitatively evaluated through a benchmark model developed specifically for this study. Compared to the benchmark model, the 'demons' registration has the maximum error of 1.2 cm, which can be corrected by the FEM to 0.4 cm, and the average error of the 'demons' registration is reduced from 0.17 to 0.11 cm. For the CT images of lung and prostate patients, the deformation maps generated by the 'demons' algorithm were found unrealistic at several places. In these places, the displacement differences between the 'demons' registrations and their FEM corrections were found in the range of 0.4 and 1.1 cm. The mesh refinement and FEM simulation were implemented in a single thread application which requires about 45 min of computation time on a 2.6 GHz computer. This study has demonstrated that the FEM can be integrated with intensity-based image registration algorithms to improve their registration accuracy, especially in low-contrast regions.     

A deformable image registration method to handle distended rectums in prostate cancer radiotherapy

In image-guided adaptive radiotherapy, it is important to have the capability to automatically and accurately delineate the rectal wall, which is a major dose-limiting organ in prostate cancer radiotherapy. As image registration is a process to find the spatial correspondence between two images, a major challenge in intensity-based deformable image registration is to deal with the situation where no correspondence exists for some objects between the two images to be registered. One example is the variation of rectal contents due to the presence and absence of bowel gas. The intensity-based deformable image registration methods alone cannot create the correct spatial transformation if there is no correspondence between the source and target images. In this study we implemented an automatic image intensity modification procedure to create artificial gas pockets in the planning computed tomography (CT) images. A diffusion-based deformable image registration algorithm was developed to use an adaptive smoothing algorithm to better handle large organ deformations. The process was tested in 15 prostate cancer cases and 30 daily CT images containing the largest distended rectums. The manually delineated rectums agreed well with the autodelineated rectums when using the image-intensity modification procedure.     

Tracking lung tissue motion and expansion/compression with inverse consistent image registration and spirometry

Breathing motion is one of the major limiting factors for reducing dose and irradiation of normal tissue for conventional conformal radiotherapy. This paper describes a relationship between tracking lung motion using spirometry data and image registration of consecutive CT image volumes collected from a multislice CT scanner over multiple breathing periods. Temporal CT sequences from 5 individuals were analyzed in this study. The couch was moved from 11 to 14 different positions to image the entire lung. At each couch position, 15 image volumes were collected over approximately 3 breathing periods. It is assumed that the expansion and contraction of lung tissue can be modeled as an elastic material. Furthermore, it is assumed that the deformation of the lung is small over one-fifth of a breathing period and therefore the motion of the lung can be adequately modeled using a small deformation linear elastic model. The small deformation inverse consistent linear elastic image registration algorithm is therefore well suited for this problem and was used to register consecutive image scans. The pointwise expansion and compression of lung tissue was measured by computing the Jacobian of the transformations used to register the images. The logarithm of the Jacobian was computed so that expansion and compression of the lung were scaled equally. The log-Jacobian was computed at each voxel in the volume to produce a map of the local expansion and compression of the lung during the breathing period. These log-Jacobian images demonstrate that the lung does not expand uniformly during the breathing period, but rather expands and contracts locally at different rates during inhalation and exhalation. The log-Jacobian numbers were averaged over a cross section of the lung to produce an estimate of the average expansion or compression from one time point to the next and compared to the air flow rate measured by spirometry. In four out of five individuals, the average log-Jacobian value and the air flow rate correlated well (R2 = 0.858 on average for the entire lung). The correlation for the fifth individual was not as good (R2 = 0.377 on average for the entire lung) and can be explained by the small variation in tidal volume for this individual. The correlation of the average log-Jacobian value and the air flow rate for images near the diaphragm correlated well in all five individuals (R2 = 0.943 on average). These preliminary results indicate a strong correlation between the expansion/compression of the lung measured by image registration and the air flow rate measured by spirometry. Predicting the location, motion, and compression/expansion of the tumor and normal tissue using image registration and spirometry could have many important benefits for radiotherapy treatment. These benefits include reducing radiation dose to normal tissue, maximizing dose to the tumor, improving patient care, reducing treatment cost, and increasing patient throughput.     

基于分段B样条形变配准方法的头颈部伪CT生成研究

目的:探讨分段B样条形变配准方法在头颈部伪CT(sCT)生成中的应用,以及对sCT生成精度的影响。方法:收集已经进行调强放射治疗的鼻咽癌患者45例,每例计划均包括头颈部T1加权核磁共振成像(MRI)和CT图像。使用3D Slicer软件对MRI和CT图像分别进行分段B样条形变配准、整体B样条形变配准、分段刚性配准和整体刚性配准4种方法配准,比较配准后的MRI图像和真实CT图像的Dice相似性系数(DSC)值。随机选取其中的30例患者作为训练集,15例患者为测试集,将配准后的MRI和CT图像通过pix2pix网络进行模型训练生成sCT,对生成的sCT和真实CT进行平均绝对误差(MAE)、结构相似性系数(SSIM)和峰值信噪比(PSNR)值的比较,分析通过阈值法分割为不同组织(骨头、软组织、空气和脂肪)的MAE值。结果:配准后的MRI和真实CT图像比较,分段B样条形变配准方法的DSC值最优;使用4种配准方法生成的sCT和真实CT图像进行MAE、SSIM和PSNR值比较,分段配准方法比整体配准方法好,B样条形变配准方法比刚性配准方法好。分段B样条形变配准方法的MAE值为(74.783±9.8...     

基于图像变形配准的肺癌自适应放疗剂量学研究

目的:应用图像配准技术实现肺癌自适应放疗中剂量的累加,并评价放疗计划中靶区、正常组织和危及器官相应的剂量学改变。方法:选取9例接受自适应调强放射治疗的肺癌患者,这些患者在经过20次分次治疗后,重新采集CT图像,运用变形图像配准技术将2次CT图像进行剂量累加,得到累加剂量以及相关计量学参数,然后比较自适应放疗及常规的调强放射的剂量学差异。结果:经自适应放疗,大体肿瘤体积(GTV)体积相对于放疗前平均缩小53.2%,靶区肿瘤受照剂量相对于常规调强放疗计划平均提高0.41 Gy;肺组织V_(20)、V_(30)分别平均降低2.17%、3.32%;心脏V_(30)平均降低1.14%,V_(40)降低2.98%;脊髓最大受照剂量降低1.21 Gy。结论:肺癌放疗过程中,自适应放疗相对于常规调强放疗能提高靶区受照剂量,有效减少周围正常组织剂量,降低放疗副作用的发生。     

Multi-dimensional respiratory motion tracking from markerless optical surface imaging based on deformable mesh registration

Real-time optical surface imaging systems offer a non-invasive way to monitor intra-fraction motion of a patient's thorax surface during radiotherapy treatments. Due to lack of point correspondence in dynamic surface acquisition, such systems cannot currently provide 3D motion tracking at specific surface landmarks, as available in optical technologies based on passive markers. We propose to apply deformable mesh registration to extract surface point trajectories from markerless optical imaging, thus yielding multi-dimensional breathing traces. The investigated approach is based on a non-rigid extension of the iterative closest point algorithm, using a locally affine regularization. The accuracy in tracking breathing motion was quantified in a group of healthy volunteers, by pair-wise registering the thoraco-abdominal surfaces acquired at three different respiratory phases using a clinically available optical system. The motion tracking accuracy proved to be maximal in the abdominal region, where breathing motion mostly occurs, with average errors of 1.09 mm. The results demonstrate the feasibility of recovering multi-dimensional breathing motion from markerless optical surface acquisitions by using the implemented deformable registration algorithm. The approach can potentially improve respiratory motion management in radiation therapy, including motion artefact reduction or tumour motion compensation by means of internal/external correlation models.     

Technical note: A novel boundary condition using contact elements for finite element based deformable image registration

Deformable image registration is an important tool for image-guided radiotherapy. Physics-model-based deformable image registration using finite element analysis is one of the methods currently being investigated. The calculation accuracy of finite element analysis is dependent on given boundary conditions, which are usually based on the surface matching of the organ in two images. Such a surface matching, however, is hard to obtain from medical images. In this study, we developed a new boundary condition to circumvent the traditional difficulties. Finite element contact-impact analysis was employed to simulate the interaction between the organ of interest and the surrounding body. The displacement loading is not necessarily specified. The algorithm automatically deforms the organ model into the minimum internal energy state. The analysis was performed on CT images of the lung at two different breathing phases (exhalation and full inhalation). The result gave the displacement vector map inside the lung. Validation of the result showed satisfactory agreement in most parts of the lung. This approach is simple, operator independent and may provide improved accuracy of the prediction of organ deformation.     

一种基于2D/3D配准的脊柱术中校正方法

背景：脊柱术前三维影像有助于诊断和治疗,术中患者体位变化将引起脊柱形态改变,致使术前影像不能反映术中实际情况,无法确保手术的顺利实施。 目的：利用脊髓手术中影像校正术前脊柱模型形态。 方法：实验提出了一种基于2D/3D配准的脊柱术中校正方法,借助数字影像重建技术完成术前X射线图像与CT体数据的2D/3D配准,进一步完成术中、术前X射线图像中独立椎段的特征匹配,利用上述配准结果实现术前脊柱CT模型的术中快速校正。 结果与结论：采用附有标记的颈椎标本进行实验,校正后可基本消除术前脊柱模型与术中形态的偏差,其误差可控制在1 mm以内,能够满足医学临床要求。     

A deformable phantom for 4D radiotherapy verification: design and image registration evaluation

Motion of thoracic tumors with respiration presents a challenge for three-dimensional (3D) conformal radiation therapy treatment. Validation of techniques aimed at measuring and minimizing the effects of respiratory motion requires a realistic deformable phantom for use as a gold standard. The purpose of this study was to develop and study the characteristics of a reproducible, tissue equivalent, deformable lung phantom. The phantom consists of a Lucite cylinder filled with water containing a latex balloon stuffed with dampened natural sponges. The balloon is attached to a piston that mimics the human diaphragm. Nylon wires and Lucite beads, emulating vascular and bronchial bifurcations, were uniformly glued at various locations throughout the sponges. The phantom is capable of simulating programmed irregular breathing patterns with varying periods and amplitudes. A tissue equivalent tumor, suitable for holding radiochromic film for dose measurements was embedded in the sponge. To assess phantom motion, eight 3D computed tomography data sets of the static phantom were acquired for eight equally spaced positions of the piston. The 3D trajectories of 12 manually chosen point landmarks and the tumor center-of-mass were studied. Motion reproducibility tests of the deformed phantom were established on seven repeat scans of three different states of compression. Deformable image registration (DIR) of the extreme breathing phases was performed. The accuracy of the DIR was evaluated by visual inspection of image overlays and quantified by the distance-to-agreement (DTA) of manually chosen point landmarks and triangulated surfaces obtained from 3D contoured structures. In initial tests of the phantom, a 20-mm excursion of the piston resulted in deformations of the balloon of 20 mm superior-inferior, 4 mm anterior-posterior, and 5 mm left-right. The change in the phantom mean lung density ranged from 0.24 (0.12 SD) g/cm3 at peak exhale to 0.19 (0.12 SD) g/cm3 at peak inhale. The SI displacement of the landmarks varied between 94% and 3% of the piston excursion for positions closer and farther away from the piston, respectively. The reproducibility of the phantom deformation was within the image resolution (0.7 x 0.7 x 1.25 mm3). Vector average registration accuracy based on point landmarks was found to be 0.5 (0.4 SD) mm. The tumor and lung mean 3D DTA obtained from triangulated surfaces were 0.4 (0.1 SD) mm and 1.0 (0.8 SD) mm, respectively. This phantom is capable of reproducibly emulating the physically realistic lung features and deformations and has a wide range of potential applications, including four-dimensional (4D) imaging, evaluation of deformable registration accuracy, 4D planning and dose delivery.     

基于混合优化算法的医学图像配准方法

基于互信息的图像配准算法计算复杂度高,配准速度慢。针对这一问题,本文提出一种基于改进遗传算法和Powell算法相结合的医学图像配准方法。首先针对传统遗传算法收敛速度慢、易早熟的缺陷,本文对遗传操作中的交叉运算过程提出了改进策略,并将改进的遗传算法与Powell算法相结合,充分利用遗传算法的全局搜索能力与Powell算法的局部搜索能力。与Powell算法和未改进的遗传算法相比,本文提出的算法极大地缩短了图像配准所用的时间,同时提高了算法的抗噪性。     

A novel magnetic resonance-positive emission image registration based on morphology

Structures were obtained from images with morphological characteristics to register with voxel-based method. We applied simple morphological operations to obtain human brain cortex and chose normalized mutual information as similarity measure for the geometric alignment of PET and MR images. Evaluation used nine patients, and the results showed that sub-voxel accuracy was achieved and the registration process was significantly more rapid. Thus this new automated multi-modality registration method is more robust and has high accuracy.     

A fast image reconstruction algorithm based on penalized-likelihood estimate

Statistical iterative methods for image reconstruction like maximum likelihood expectation maximization (ML-EM) are more robust and flexible than analytical inversion methods and allow for accurately modeling the counting statistics and the photon transport during acquisition. They are rapidly becoming the standard for image reconstruction in emission computed tomography. The maximum likelihood approach provides images with superior noise characteristics compared to the conventional filtered back projection algorithm. But a major drawback of the statistical iterative image reconstruction is its high computational cost. In this paper, a fast algorithm is proposed as a modified OS-EM (MOS-EM) using a penalized function, which is applied to the least squares merit function to accelerate image reconstruction and to achieve better convergence. The experimental results show that the algorithm can provide high quality reconstructed images with a small number of iterations.