Enhancing the performance of model-based elastography by incorporating additional a priori information in the modulus image reconstruction process

Model-based elastography is fraught with problems owing to the ill-posed nature of the inverse elasticity problem. To overcome this limitation, we have recently developed a novel inversion scheme that incorporates a priori information concerning the mechanical properties of the underlying tissue structures, and the variance incurred during displacement estimation in the modulus image reconstruction process. The information was procured by employing standard strain imaging methodology, and introduced in the reconstruction process through the generalized Tikhonov approach. In this paper, we report the results of experiments conducted on gelatin phantoms to evaluate the performance of modulus elastograms computed with the generalized Tikhonov (GTK) estimation criterion relative to those computed by employing the un-weighted least-squares estimation criterion, the weighted least-squares estimation criterion and the standard Tikhonov method (i.e., the generalized Tikhonov method with no modulus prior). The results indicate that modulus elastograms computed with the generalized Tikhonov approach had superior elastographic contrast discrimination and contrast recovery. In addition, image reconstruction was more resilient to structural decorrelation noise when additional constraints were imposed on the reconstruction process through the GTK method.     

The importance and implementation of accurate 3D compensation methods for quantitative SPECT

The purpose of this study was to investigate the importance of 2D versus 3D compensation methods in SPECT. The compensation methods included in the study addressed two important degrading factors, namely attenuating and collimator-detector response in SPET. They can be divided into two general categories. The conventional methods are based on the filtered backprojection algorithm, the Chang algorithm for attenuation compensation and the Metz filter for detector response compensation. These methods, which were computationally efficient, could only achieve approximate compensation due to the assumptions made. The quantitative compensation methods provide accurate compensation by modelling the degrading effects at the expense of large computational requirements. Both types of compensation methods were implemented in 2D and 3D reconstructions. The 2D and 3D reconstruction/compensation methods were evaluated using data from simulation of brain and heart, and patient thallium SPECT studies. Our results demonstrate the importance of compensation methods in improving the quality and quantitative accuracy of SPECT images and the relative effectiveness of the different 2D and 3D reconstruction/compensation methods. We concluded that 3D implementation of the quantitative compensation methods provides the best SPECT image in terms of quantitative accuracy, spatial resolution, and noise at a cost of high computational requirements.     

Use of a priori information in estimating tissue resistivities--application to measured data.

A statistically constrained minimum mean squares error estimator (MiMSEE) has been shown to be useful in estimating internal resistivity distribution by the use of simulated data. In this study, the performance of the MiMSEE algorithm is tested by using measured data from resistor phantoms. The MiMSEE uses a priori information on body geometry, electrode position, statistical properties of tissue resistivities, instrumentation noise and linearization error to calculate the optimum inverse matrix which maps the surface potentials to unknown regional resistivities. In this study, the MiMSEE is also constrained with the variance-covariance of the modelling error to improve the estimation accuracy. The data are obtained from two different phantom geometries, namely five-region and thorax. Using the measured data, the estimations are realized and errors are calculated. Then, the results are compared with the results obtained by using a conventional least squares error estimator (LSEE). The five-region model results show similarity with the simulation study results of Baysal and Eyübo?lu. On the thorax model, the total estimation error is 34.2% with the MiMSEE compared with 856% with the LSEE. It is concluded that the MiMSEE is more robust than the LSEE and applicable to measured data.     

Comparative evaluation of image segmentation methods for volume quantitation in SPECT.

A wide variety of image segmentation techniques have been proposed for the measurement of organ or lesion volumes in SPECT images. Evaluation of the relative performance of the various methods is difficult due to wide variations in system response characteristics, size, shape, and contrast of the imaged objects, and image acquisition and processing techniques. Selected image segmentation methods for volume quantitation in SPECT were applied to a set of simulated SPECT images containing objects ranging in volume from 1.8 to 113.1 cc. The specific segmentation methods included: (1) operator drawn regions of interest, (2) count-based methods, (3) three levels of fixed thresholds, (4) an adaptive threshold (GLH method), (5) a two-dimensional (2-D) edge detection method, and (6) a three-dimensional (3-D) edge detection method. In general, the 3-D edge detection method required minimal operator intervention while providing the most accurate and consistent estimates of object volume across changes in object contrast and size.     

Image analysis methods for assessing levels of image plane nonuniformity and stochastic noise in a magnetic resonance image of a homogeneous phantom

Magnetic response image plane nonuniformity and stochastic noise are properties that greatly influence the outcome of quantitative magnetic resonance imaging (MRI) evaluations such as gel dosimetry measurements using MRI. To study these properties, robust and accurate image analysis methods are required. New nonuniformity level assessment methods were designed, since previous methods were found to be insufficiently robust and accurate. The new and previously reported nonuniformity level assessment methods were analyzed with respect to, for example, insensitivity to stochastic noise; and previously reported stochastic noise level assessment methods with respect to insensitivity to nonuniformity. Using the same image data, different methods were found to assess significantly different levels of nonuniformity. Nonuniformity levels obtained using methods that count pixels in an intensity interval, and obtained using methods that use only intensity values, were found not to be comparable. The latter were found preferable, since they assess the quantity intrinsically sought. A new method which calculates a deviation image, with every pixel representing the deviation from a reference intensity, was least sensitive to stochastic noise. Furthermore, unlike any other analyzed method, it includes all intensity variations across the phantom area and allows for studies of nonuniformity shapes. This new method was designed for accurate studies of nonuniformities in gel dosimetry measurements, but could also be used with benefit in quality assurance and acceptance testing of MRI, scintillation camera, and computer tomography systems. The stochastic noise level was found to be greatly method dependent. Two methods were found to be insensitive to nonuniformity and also simple to use in practice. One method assesses the stochastic noise level as the average of the levels at five different positions within the phantom area, and the other assesses the stochastic noise in a region outside the phantom area.     

The importance of the accuracy of image registration of SPECT images for 3D targeted radionuclide therapy dosimetry

In this paper, the importance of the accuracy of image registration of time-sequential SPECT images for 3D targeted radionuclide therapy dosimetry is studied. Image registration of a series of SPECT scans is required to allow the computation of the 3D absorbed dose distribution for both tumour sites and normal organs. Three simulated 4D datasets, based on patient therapy studies, were generated to allow the effect of mis-registration on the absorbed dose distribution to be investigated. The tumour sites studied range in size, shape and position, relative to the centre of the 3D SPECT scan. Randomly generated transformations along the x-, y- and z-axes and rotations around the z-axis were employed and the maximum and average absorbed dose distribution statistics, for the tumour sites present, were computed. It was shown that even small mis-registrations, translation of less than 9 mm and rotation of less than 5 degrees might cause differences in the absorbed dose statistics of up to 90%, especially when the size of the tumour is comparable to the induced mis-registration or when the tumour is situated close to the edge of the 3D dataset.     

单光子发射型计算机断层成像与CT融合图像评价125I粒子在靶器官中的分布

目的 结合全身扫描,探讨单光子发射型计算机断层成像(SPECT)与CT融合图像对植入的放射性125I粒子在体内分布的评价.方法 2008年5月至2009年6月复旦大学附属中山医院核医学科接受治疗的47例125I粒子植入患者中,6例前列腺癌患者实施了超声诱导下125I粒子植入,另41例有明确肿瘤病史并伴有门静脉或上腔静脉癌栓患者,在数字减影血管成像(DSA)诱导下行支架和125I粒子链条植入到血管腔内.所有患者在125I粒子植入24 h后行全身扫描,再根据125I粒子所在部位行SPECT与CT显像检查.结合全身扫描图像和SPECT与CT图像对植入125I粒子的分布情况进行评价.结果 全身扫描图像为SPECT与CT检查提供准确定位,但无法显示粒子的准确解剖部位.6例前列腺癌患者中,SPECT与CT融合图像显示5例患者植入粒子在前列腺内分布较好,而1例患者粒子植入到右侧精囊腺内;41例有明确肿瘤病史并伴有门静脉或上腔静脉癌栓患者中,SPECT与CT融合图像显示40例血管腔内的粒子链条位于合理的位置,1例植入到上腔静脉内的粒子链条脱离到右心腔内.结论 结合全身扫描,SPECT与CT融合图像可以评价125I粒子在靶器官的分布情况,能及时发现位置偏离的粒子.     

单光子发射型计算机断层成像与CT融合图像评价^125I粒子在靶器官中的分布

目的结合全身扫描,探讨单光子发射型计算机断层成像（SPECT）与CT融合图像对植入的放射性^125I粒子在体内分布的评价。方法2008年5月至2009年6月复旦大学附属中山医院核医学科接受治疗的47例^125I粒子植入患者中,6例前列腺癌患者实施了超声诱导下^125I粒子植入,另41例有明确肿瘤病史并伴有门静脉或上腔静脉癌栓患者,在数字减影血管成像（DSA）诱导下行支架和^125I粒子链条植入到血管腔内。所有患者在^125I粒子植入24h后行全身扫描,再根据^125I粒子所在部位行SPECT与CT显像检查。结合全身扫描图像和SPECT与CT图像对植入^125I粒子的分布情况进行评价。结果全身扫描图像为SPECT与CT检查提供准确定位,但无法显示粒子的准确解剖部位。6例前列腺癌患者中,SPECT与CT融合图像显示5例患者植入粒子在前列腺内分布较好,而1例患者粒子植入到右侧精囊腺内;41例有明确肿瘤病史并伴有门静脉或上腔静脉癌栓患者中,SPECT与CT融合图像显示40例血管腔内的粒子链条位于合理的位置,1例植入到上腔静脉内的粒子链条脱离到右心腔内。结论结合全身扫描,SPECT与CT融合图像可以评价^125I粒子在靶器官的分布情况,能及时发现位置偏离的粒子。     

Use of 3D reconstruction to correct for patient motion in SPECT

Patient motion occurring during data acquisition in single photon emission computed tomography (SPET) can cause serious reconstruction artefacts. We have developed a new approach to correct for head motion in brain SPECT. Prior to motion, projections are assigned to conventional projections. When head motion occurs, it is measured by a motion monitoring system, and subsequent projection data are mapped to 'virtual' projections. The appropriate position of each virtual projection is determined by applying the converse of the patient's accumulated motion to the actual camera projection. Conventional and virtual projections, taken together, form a consistent set that can be reconstructed using a three-dimensional (3D) algorithm. The technique has been tested on a range of simulated rotational movements, both within and out of the transaxial plane. For all simulated movements, the motion corrected images exhibited better agreement with a motion free reconstruction than did the uncorrected images. This technique may help to overcome one of the major remaining limitations on image quality and quantitative accuracy in SPECT.     

Inversion of the 3D Radon transform for a multidetector, point-focused SPECT brain scanner

The Radon transform is presented for unattenuated projection data acquired with a multidetector, point-focused SPECT brain scanner. The three-dimensional (3D) integral transform is shown to be exactly invertible only for the case of an ideal machine whose detectors scan away from the source to infinity in all directions. The two-dimensional (2D) ramp filter reconstruction algorithm previously used for this scanner is a limiting case of the 3D method and is exact only for sources with cylindrical symmetry, (f(x,y,z) = f(x,y) for all z). Projection data from a long cylinder and a 1.5 cm thick disc of equal diameters and activity concentrations were simulated by computer for the same scan pattern used on the machine. The data were reconstructed with both the 2D and 3D analytic methods. The 2D method produced a 20% bowl-shaped dip in the centre of the disc, whereas the central slice of the 3D reconstruction was more than 97% accurate. The effective, noise-equivalent sensitivity when reconstructing the central slice of the disc with the necessary 3D method is 10.1 times lower than the sensitivity obtained for the long cylindrical source when reconstructing with the 2D method.     

医学影像导出三维模型并建立3D PDF文件的方法

目的探讨医学影像导出为三维模型并建立交互3D PDF文件。方法使用了3个开源软件包括3D Slicer、Mesh Lab和Te Xstudio,进行感兴趣区体裁剪、模型分割和3D PDF制作。结果 3D Slicer可以多种方法进行模型分割,Mesh Lab可以进行模型清理和分割,并在导出U3D格式文件的同时生成TEX格式文件,通过Te Xstudio编译生成3D PDF文件。结论这种方法步骤简单易学,推广普及容易,有利于促进更多3D PDF的制作和使用。     

The Monte Carlo SRNA-VOX code for 3D proton dose distribution in voxelized geometry using CT data

This paper describes the application of the SRNA Monte Carlo package for proton transport simulations in complex geometry and different material compositions. The SRNA package was developed for 3D dose distribution calculation in proton therapy and dosimetry and it was based on the theory of multiple scattering. The decay of proton induced compound nuclei was simulated by the Russian MSDM model and our own using ICRU 63 data. The developed package consists of two codes: the SRNA-2KG, which simulates proton transport in combinatorial geometry and the SRNA-VOX, which uses the voxelized geometry using the CT data and conversion of the Hounsfield's data to tissue elemental composition. Transition probabilities for both codes are prepared by the SRNADAT code. The simulation of the proton beam characterization by multi-layer Faraday cup, spatial distribution of positron emitters obtained by the SRNA-2KG code and intercomparison of computational codes in radiation dosimetry, indicate immediate application of the Monte Carlo techniques in clinical practice. In this paper, we briefly present the physical model implemented in the SRNA package, the ISTAR proton dose planning software, as well as the results of the numerical experiments with proton beams to obtain 3D dose distribution in the eye and breast tumour.     

Comparison of methods for suppressing edge and aliasing artefacts in iterative x-ray CT reconstruction

X-ray CT images obtained with iterative reconstruction (IR) can be hampered by the so-called edge and aliasing artefacts, which appear as interference patterns and severe overshoots in the areas of sharp intensity transitions. Previously, we have demonstrated that these artefacts are caused by discretization errors during the projection simulation step in IR. Although these errors are inherent to IR, they can be adequately suppressed by reconstruction on an image grid that is finer than that typically used for analytical methods such as filtered back-projection. Two other methods that may prevent edge artefacts are: (i) smoothing the projections prior to reconstruction or (ii) using an image representation different from voxels; spherically symmetric Kaiser-Bessel functions are a frequently employed example of such a representation. In this paper, we compare reconstruction on a fine grid with the two above-mentioned alternative strategies for edge artefact reduction. We show that the use of a fine grid results in a more adequate suppression of artefacts than the smoothing of projections or using the Kaiser-Bessel image representation.     

Statistical tracking of tree-like tubular structures with efficient branching detection in 3D medical image data

The segmentation of tree-like tubular structures such as coronary arteries and airways is an essential step for many 3D medical imaging applications. Statistical tracking techniques for the extraction of elongated structures have received considerable attention in recent years due to their robustness against image noise and pathological changes. However, most tracking methods are limited to a specific application and do not support branching structures efficiently. In this work, we present a novel statistical tracking approach for the extraction of different types of tubular structures with ringlike cross-sections. Domain-specific knowledge is learned from training data sets and integrated into the tracking process by simple adaption of parameters. In addition, an efficient branching detection algorithm is presented. This approach was evaluated by extracting coronary arteries from 32 CTA data sets and distal airways from 20 CT scans. These data sets were provided by the organizers of the workshop '3D Segmentation in the Clinic: A Grand Challenge II-Coronary Artery Tracking (CAT08)' and 'Extraction of Airways from CT 2009 (EXACT'09)'. On average, 81.5% overlap and 0.51 mm accuracy for the tracking of coronary arteries were achieved. For the extraction of airway trees, 51.3% of the total tree length, 53.6% of the total number of branches and a 4.98% false positive rate were attained. In both experiments, our approach is comparable to state-of-the-art methods.     

A 3D model of non-uniform attenuation and detector response for efficient iterative reconstruction in SPECT

A 3D physical model for iterative reconstruction in SPECT has been developed and applied to experimental data. The model incorporates non-uniform attenuation using reconstructed transmission CT data and distance-dependent detector response based on response function measurements over a range of distances from the detector. The 3D model has been implemented in a computationally efficient manner with practical memory requirements. The features of the model that provide efficiency are described including a new region-dependent reconstruction (RDR) technique. With RDR, filtered backprojection is used to reconstruct areas of the image of minimal clinical importance, and the result is used to supplement the iterative reconstruction of the clinically important areas of the image. The 3D model was incorporated into the maximum likelihood-expectation maximization (ML-EM) reconstruction algorithm and tested in three phantom studies--a point source, a uniform cylinder, and an anthropomorphic thorax--and a patient 9Tc(m) sestamibi study. Reconstructed images with the 3D method exhibited excellent noise and resolution characteristics. With the sestamibi data, the RDR technique produced essentially the conventional ML-EM estimate in the cardiac region with substantial time savings.     

4D medical image computing and visualization of lung tumor mobility in spatio-temporal CT image data

The development of 4D CT imaging has introduced the possibility of measuring breathing motion of tumors and inner organs. Conformal thoracic radiation therapy relies on a quantitative understanding of the position of lungs, lung tumors, and other organs during radiation delivery. Using 4D CT data sets, medical image computing and visualization methods were developed to visualize different aspects of lung and lung tumor mobility during the breathing cycle and to extract quantitative motion parameters. A non-linear registration method was applied to estimate the three-dimensional motion field and to compute 3D point trajectories. Specific visualization techniques were used to display the resulting motion field, the tumor's appearance probabilities during a breathing cycle as well as the volume covered by the moving tumor. Furthermore, trajectories of the tumor center-of-mass and organ specific landmarks were computed for the quantitative analysis of tumor and organ motion. The analysis of 4D data sets of seven patients showed that tumor mobility differs significantly between the patients depending on the individual breathing pattern and tumor location.     

Noninvasive image analysis of 3D construct mineralization in a perfusion bioreactor

Although the beneficial effects of perfusion on cell-mediated mineralization have been demonstrated in several studies, the size of the mineralized constructs produced has been limited. The ability to quantify mineralized matrix formation non-invasively within 3D constructs would benefit efforts to optimize bioreactor conditions for scaling-up constructs to clinically relevant dimensions. In this study, we report a micro-CT imaging-based technique to monitor 3D mineralization over time in a perfusion bioreactor and specifically assess mechanisms of construct mineralization by quantifying the number, size, and distribution of mineralized particle formation within constructs varying in thickness from 3 to 9 mm. As expected, mineralized matrix volume and particle number increased with construct thickness. Analyzing multiple concentric volumes inside each construct indicated that a greater proportion of the mineral volume was found within the interior of the perfused constructs. Interestingly, intermediate-sized 6mm thick constructs were found to have the highest core mineral volume fraction and the largest mineralized particles. Two complementary mechanisms of increasing total mineral volume were observed in the 6 and 9 mm constructs: increasing particle size and increasing the number of mineralized particles, respectively. The rate of mineralized matrix formation in the perfused constructs increased from 0.69 mm(3)/week during the first 3 weeks of culture to 1.03 mm(3)/week over the final 2 weeks. In contrast, the rate of mineral deposition in the static controls was 0.01 mm(3)/week during the first 3 weeks of culture and 0.16 mm(3)/week from week 3 to week 5. The ability to monitor overall construct mineralization non-invasively coupled with quantitative analysis of mineralized particle size, number, and distribution offers a powerful tool for elucidating how mineral growth mechanisms are affected by cell type, scaffold material and architecture, or bioreactor flow conditions.     

Effects of x-ray and CT image enhancements on the robustness and accuracy of a rigid 3D/2D image registration

A rigid body three-dimensional/two-dimensional (3D/2D) registration method has been implemented using mutual information, gradient ascent, and 3D texturemap-based digitally reconstructed radiographs. Nine combinations of commonly used x-ray and computed tomography (CT) image enhancement methods, including window leveling, histogram equalization, and adaptive histogram equalization, were examined to assess their effects on accuracy and robustness of the registration method. From a set of experiments using an anthropomorphic chest phantom, we were able to draw several conclusions. First, the CT and x-ray preprocessing combination with the widest attraction range was the one that linearly stretched the histograms onto the entire display range on both CT and x-ray images. The average attraction ranges of this combination were 71.3 mm and 61.3 deg in the translation and rotation dimensions, respectively, and the average errors were 0.12 deg and 0.47 mm. Second, the combination of the CT image with tissue and bone information and the x-ray images with adaptive histogram equalization also showed subvoxel accuracy, especially the best in the translation dimensions. However, its attraction ranges were the smallest among the examined combinations (on average 36 mm and 19 deg). Last the bone-only information on the CT image did not show convergency property to the correct registration.     

The cost of quality: Implementing generalization and suppression for anonymizing biomedical data with minimal information loss

ObjectiveWith the ARX data anonymization tool structured biomedical data can be de-identified using syntactic privacy models, such as k-anonymity. Data is transformed with two methods: (a) generalization of attribute values, followed by (b) suppression of data records. The former method results in data that is well suited for analyses by epidemiologists, while the latter method significantly reduces loss of information. Our tool uses an optimal anonymization algorithm that maximizes output utility according to a given measure. To achieve scalability, existing optimal anonymization algorithms exclude parts of the search space by predicting the outcome of data transformations regarding privacy and utility without explicitly applying them to the input dataset. These optimizations cannot be used if data is transformed with generalization and suppression. As optimal data utility and scalability are important for anonymizing biomedical data, we had to develop a novel method.MethodsIn this article, we first confirm experimentally that combining generalization with suppression significantly increases data utility. Next, we proof that, within this coding model, the outcome of data transformations regarding privacy and utility cannot be predicted. As a consequence, existing algorithms fail to deliver optimal data utility. We confirm this finding experimentally. The limitation of previous work can be overcome at the cost of increased computational complexity. However, scalability is important for anonymizing data with user feedback. Consequently, we identify properties of datasets that may be predicted in our context and propose a novel and efficient algorithm. Finally, we evaluate our solution with multiple datasets and privacy models.ResultsThis work presents the first thorough investigation of which properties of datasets can be predicted when data is anonymized with generalization and suppression. Our novel approach adopts existing optimization strategies to our context and combines different search methods. The experiments show that our method is able to efficiently solve a broad spectrum of anonymization problems.ConclusionOur work shows that implementing syntactic privacy models is challenging and that existing algorithms are not well suited for anonymizing data with transformation models which are more complex than generalization alone. As such models have been recommended for use in the biomedical domain, our results are of general relevance for de-identifying structured biomedical data.