Improving accuracy and efficiency of mutual information for multi-modal retinal image registration using adaptive probability density estimation

Mutual information (MI) is a popular similarity measure for performing image registration between different modalities. MI makes a statistical comparison between two images by computing the entropy from the probability distribution of the data. Therefore, to obtain an accurate registration it is important to have an accurate estimation of the true underlying probability distribution. Within the statistics literature, many methods have been proposed for finding the ‘optimal’ probability density, with the aim of improving the estimation by means of optimal histogram bin size selection. This provokes the common question of how many bins should actually be used when constructing a histogram. There is no definitive answer to this. This question itself has received little attention in the MI literature, and yet this issue is critical to the effectiveness of the algorithm. The purpose of this paper is to highlight this fundamental element of the MI algorithm. We present a comprehensive study that introduces methods from statistics literature and incorporates these for image registration. We demonstrate this work for registration of multi-modal retinal images: colour fundus photographs and scanning laser ophthalmoscope images. The registration of these modalities offers significant enhancement to early glaucoma detection, however traditional registration techniques fail to perform sufficiently well. We find that adaptive probability density estimation heavily impacts on registration accuracy and runtime, improving over traditional binning techniques.     

Computer-aided intrapatient comparison of brain SPECT images: the gray-level normalization issue applied to children with epilepsy.

A tool was developed for automated intrapatient comparison of brain SPECT images, with specific emphasis on gray-level normalization. METHODS: Ictal and interictal (99m)Tc-ethyl cysteinate dimer SPECT images were acquired for 6 children with partial epilepsy (age range, 2-10 y). For each patient, 3-dimensional rigid geometric ictal-to-interictal image registration optimizing different classic criteria (correlation coefficient, ratio uniformity) in a multiscale translation-rotation 6-parameter space was first performed. Gray-level normalization was then performed with different methods, using a 1- or 2-parameter linear model. In the 1-parameter case, the scaling factor was equal to the interictal-to-ictal ratio of the maximum, mean, or median values calculated within different reference volumes (whole brain or cerebellum) or obtained by linear regression between ictal and interictal counts in the brain or by maximizing a robust criterion, the number of deterministic sign changes in the subtraction images. In the 2-parameter case, the scaling factor and additive constant were estimated using these last 2 methods. For each patient, registration validity and normalization plausibility were assessed by considering the correlation scatterplot together with the different normalization lines and by comparing interictal and registered normalized ictal images using a twin display (with isocontours) in the 3 orthogonal planes. Three-dimensional volumes of interest could be selected on coupled interictal-subtraction images for further focused numeric comparison. RESULTS: After a satisfactory and stable geometric registration with both criteria, the different normalization methods led to similar subtraction images for 5 of 6 patients, except the maxima ratio, which gave noticeably different results in 2 patients. For the remaining patient, with highly dissimilar ictal-interictal images, the maxima ratio normalization was obviously wrong and the other 1-parameter methods probably better depicted the data than did the 2-parameter methods. CONCLUSION: When comparing intrapatient brain SPECT images, one should be aware of the potential impact of the gray-level normalization method on clinical interpretation. For ictal-interictal images, simple robust scaling should be recommended. In particular, image maximum should generally not be considered a valid reference, and no additive constant is needed in the linear gray-level normalization model.     

全局异常信号环境下基于体素灰度多模医学图像配准研究

目的 在全局异常信号环境下,找出适合于临床应用的、满足精度和鲁棒性要求的基于体素灰度多模医学图像配准相似性测度。方法 结合对各种相似性测度的分析,对无异常信号的实际医学图像,和分别加了随机噪声及全局异常信号的多模医学图像进行配准精度的分析。结果 对各种已有成熟的相似性测度进行理论分析和实验对比研究的基础上,归一化互信息在全局异常信号环境下对多模医学图像进行配准,它们的配准精度和鲁棒性表现都令人满意,能得到准确的配准结果。而基于相关比和互信息的配准方法,不能准确地配准加了全局异常信号的多模医学图像。结论 相比于其他相似性测度,归一化互信息在全局异常信号环境下,是一个能满足配准精度和鲁棒性要求的合适相似性测度。     

基于体素灰度三维多模医学图像配准中相似性测度的选取

目的：在基于体素灰度医学图像配准领域，找出最适合于临床应用的多模医学图像配准相似性测度。方法：在极端的刚体配准条件下，检验出互相关系数，互信息和相关比相似性测度为适合的相似性测度。同时进一步解释了基于互信息相似性测度的医学图像配准易于陷入局部最优，而基于相关比相似性测度的方法易于保证配准得到全局最优，最后，利用加速的多分辨率配准方案和Powell‘s优化算法，对临床医学图像进行了基于相关比相似性测度的多模图像配准试验。结果：通过临床医学专家的判断，利用相关比相似性测度进行多模医学图像配准，安全能满足临床的要求，进行MR／CT，MR／PET三维多模医学图像配准时效果非常理想，结论：相比于其他相似性测度，互相关比相似性测度在基于体素灰度，三维多模医学图像配准领域，是一个更为适宜和准确的相似性测度。     

Fluorescein angiography and light microscopy studies of retinas irradiated by oxygen nuclei.

Fluorescein angiography and light microscopy techniques were utilized in the rhesus monkey to evaluate retinal damage following oxygen ion irradiation. Particle flux ranged from 1.3 x 10(7) particles/cm2 to 1.5 x 10(8) particles/cm2. Earliest evidence of damage was detected with fluorescein angiograms 24 h after exposures to 7.7 x 10(7) particles/cm2. No evidence of the hypothesized microlesion was found by either technique.     

Correction of motion artifacts from cardiac cine magnetic resonance images

RATIONALE AND OBJECTIVES: An image registration method was developed to automatically correct motion artifacts, mostly from breathing, from cardiac cine magnetic resonance (MR) images. MATERIALS AND METHODS: The location of each slice in an image stack was optimized by maximizing a similarity measure of the slice with another image slice stack. The optimization was performed iteratively and both image stacks were corrected simultaneously. Two procedures to optimize the similarity were tested: standard gradient optimization and stochastic optimization in which one slice is chosen randomly from the image stacks and its location is optimized. In this work, cine short- and long-axis images were used. In addition to visual inspection results from real data, the performance of the algorithm was evaluated quantitatively by simulating the movements in four real MR data sets. The mean error and standard deviation were defined for 50 simulated movements as each slice was randomly displaced. The error rate, defined as the percentage of non-satisfactory registration results, was evaluated. The paired t-test was used to evaluate the statistical difference between the tested optimization methods. RESULTS: The algorithm developed was successfully applied to correct motion artifacts from real and simulated data. The results, where typical motion artifacts were simulated, indicated an error rate of about 3%. Subvoxel registration accuracy was also achieved. When different optimization methods were compared, the registration accuracy of the stochastic approach proved to be superior to the standard gradient technique (P < 10(-9)). CONCLUSIONS: The novel method was capable of robustly and accurately correcting motion artifacts from cardiac cine MR images.     

Assessment of a technique for 2D-3D registration of cerebral intra-arterial angiography

This study assesses the ability of a computer algorithm to perform automated 2D-3D registrations of digitally subtracted cerebral angiograms. The technique was tested on clinical studies of five patients with intracranial aneurysms. The automated procedure was compared against a gold standard manual registration, and achieved a mean registration accuracy of 1.3 mm (SD 0.6 mm). Two registration strategies were tested using coarse (128 x 128 pixel) or fine (256 x 256 pixel) images. The mean registration errors proved similar but registration of the lower resolution images was 3 times quicker (mean registration times 33 s, SD 13 s for low and 150 s SD 48 s for high resolution images). The automated techniques were considerably faster than manual registrations but achieved similar accuracy. The technique has several potential uses but is particularly applicable to endovascular treatment techniques.     

Proficiency tests in the determination of activity of radionuclides in radiopharmaceutical products measured by nuclear medicine services in 8 years of comparison programmes in Brazil

Proficiency tests were applied to assess the performance of 74 nuclear medicine services in activity measurements of ¹³¹I, ¹²³I, ⁹⁹Tc^m, ⁶⁷Ga and ²⁰¹Tl. These tests produced 913 data sets from comparison programmes promoted by the National Laboratory for Ionizing Radiation Metrology (LNMRI) from 1999 to 2006.

The data were evaluated against acceptance criteria for accuracy and precision and assigned as Acceptable or Not acceptable accordingly. In addition, three other statistical parameters were used as complementary information for performance evaluation which related to normative requirements and to radionuclide calibrators.

The results have shown a necessity to improve quality control procedures and unsatisfactory performances of radionuclide calibrators, which incorporate Geiger–Müller detectors.     

基于梯度矢量流与粒子群优化算法的多模态医学图像配准

目的研究用梯度矢量流与粒子群优化算法实现多模态医学图像配准,提高配准的精度。方法算法对图像配准的特征空间、相似性测度、搜索策略3个方面进行改进：先由原始图像产生梯度矢量流场,作为配准的特征空间;然后提出并计算3种基于梯度矢量流场的相似性测度;最后使用结合了遗传算法交叉机制的粒子群优化算法找到两幅图像的最优变换。结果对仿真及实际医学图像的54次配准实验,表明该方法配准精度优于基于像素的粒子群优化方法和Walsh变换法。结论基于梯度矢量流与粒子群优化算法的图像配准方法能有效地实现多模态医学图像的配准。     

PET与MRI三维脑图像高精度配准

目的 在PC机上实现高精度的PET与MRI三维脑图像配准。方法 采用最大互信息法对6例患者PET和MRI三维脑图像进行刚体配准。使用归一化互信息作为相似性量度。在互信息计算过程中，使用Powell多参数优化法和Brent一维搜索算法。为加快配准速度，使用了多分辨金字塔方法。采用基于坐标的阈值选取方法对PET图像进行分割预处理，消除星状背景伪影。结果 配准误差平均值为2．6mm，误差中位数平均为2．7mm。结论 配准视觉效果良好，评估证明该算法可达亚体元级配准精度。     

纠正弥散张量成像图像变形的一种方法

目的：采用图像变形配准的方法校正磁共振弥散张量成像（DTI）的几何畸变，以利于神经外科导航中DTI与原始图像的融合。材料和方法：采用仿射变换和B样条的变形配准方法来矫正几何畸变，配准的相似性测度采用互信息为准则。结果：弥散张量成像特定组织间由磁敏感性差异引起的几何畸变得到了一定纠正。结论：采用3D配准的方法可以对弥散张量成像的几何畸变进行纠正，改善图像质量，提高其在神经外科导航中的临床应用价值。     

Towards efficient registration of medical images.

In this paper we propose a Bayesian based mutual information technique for image registration, combined with an established affine transformation model. Classical affine models allow the images to be approximately aligned. However, inefficiency and inaccuracy has appeared when using these affine models in rigorous circumstances, such as low-resolution images. To challenge this problem, we conduct mutual information measures with importance sampling to the images in an attempt to simulate the probability distribution of intensity similarity across the images. The entire registration adopts a stopping criterion as discovered in the context of differential equations. Finally, experimental results demonstrate the favorable performance of the proposed algorithm.     

Myocardial delineation via registration in a polar coordinate system

Rationale and Objectives. Cardiovascular disease is the number one cause of premature death in the western world. Analysis of cardiac function provides clinically useful diagnostic and prognostic information; however, manual analysis of function via delineation is prohibitively time consuming. This article describes a technique for analysis of dynamic magnetic resonance images of the left ventricle using a non-rigid registration algorithm. A manually delineated contour of a single phase was propagated through the dynamic sequence.

Materials and Methods. Short-axis cine magnetic resonance images were resampled into polar coordinates before all the time frames were aligned using a non-rigid registration algorithm. The technique was tested on 10 patient data sets, a total of 1,052 images were analyzed.

Results. Results of this approach were investigated and compared with manual delineation at all phases in the cardiac cycle, and with registration performed in a Cartesian coordinate system. The results correlated very well with manually delineated contours.

Conclusion. A novel approach to the registration and subsequent delineation of cardiac magnetic resonance images has been introduced. For the endocardium, the polar resampling technique correlated well with manual delineation, and better than for images registered without radial resampling in a Cartesian coordinate system. For the epicardium, the difference was not as apparent with both techniques correlating well.     

Electronic portal imaging device (EPID) portal image filtering for simplifying registration on radiation therapy

A simple method for improving the quality of electronic portal imaging device (EPID) portal images was proposed for the reduction of the burden on the registration between digital reconstruction radiography (DRR) and EPID portal images in radiation therapy. Conventional image filtering techniques in the spatial-frequency domain are applied to the proposed method. While a band-pass filter (BPF) is employed to extract spatial-frequency components included in the bone edge, a high-pass filter (HPF) is employed to obtain the effect corresponding to the general dynamic range compression. The band-pass filtered image is weighted by a parameter for adjusting the bone edge enhancement, and is added to the high-pass filtered image. This method was applied to the portal images in the neck region. In the image obtained by the proposed filtering, the bone edge was clearly observed. In addition, soft tissue structures were identified in the same display settings (window level/width; WL/WW) as the bone edge observation; that is, the adjustment of the display settings was not required for the observation of each object. These results suggested that both bone edge enhancement and dynamic range compression would be achieved successfully. It was estimated that the images obtained by the proposed method were more appropriate for the registration than conventional portal images, in 47 times registrations of 50 times in total (the registrations by five radiological technologists in ten patients). The proposed method was concluded to be useful for improving the quality of portal images, enabling the efficient registration.     

Automated comparison of scintigraphic images

New algorithms for automated comparison of scintigraphic images have been developed and are described here. The first step presented here is the registration of the images, performed by optimizing, with respect to the registration parameters (two translational shifts, one angle of rotation, the two parameters of a linear transformation of the gray levels), the stochastic sign change (SSC) criterion. The optimization of this criterion is demonstrated to be efficiently performed using the adaptative random search strategy; a more limited but less time-consuming method is also presented. The second step described is the point-by-point comparison of the registered images. The pixel-by-pixel application of Poisson variable statistical tests permits the generation of the significant image differences. From such images it is possible to detect modifications which escape visual inspection. Examples of applications are given in controlled and routine conditions. These algorithms are useful for the processing of many investigations and are proposed for implementation on all nuclear medicine data processing systems.     

Intensity-based 2D-3D spine image registration incorporating a single fiducial marker

RATIONALE AND OBJECTIVES: The two-dimensional (2D)-three dimensional (3D) registration of a computed tomography image to one or more x-ray projection images has a number of image-guided therapy applications. In general, fiducial marker-based methods are fast, accurate, and robust, but marker implantation is not always possible, often is considered too invasive to be clinically acceptable, and entails risk. There also is the unresolved issue of whether it is acceptable to leave markers permanently implanted. Intensity-based registration methods do not require the use of markers and can be automated because such geometric features as points and surfaces do not need to be segmented from the images. However, for spine images, intensity-based methods are susceptible to local optima in the cost function and thus need initial transformations that are close to the correct transformation. MATERIALS AND METHODS: In this report, we propose a hybrid similarity measure for 2D-3D registration that is a weighted combination of an intensity-based similarity measure (mutual information) and a point-based measure using one fiducial marker. We evaluate its registration accuracy and robustness by using gold-standard clinical spine image data from four patients. RESULTS: Mean registration errors for successful registrations for the four patients were 1.3 and 1.1 mm for the intensity-based and hybrid similarity measures, respectively. Whereas the percentage of successful intensity-based registrations (registration error < 2.5 mm) decreased rapidly as the initial transformation got further from the correct transformation, the incorporation of a single marker produced successful registrations more than 99% of the time independent of the initial transformation. CONCLUSION: The use of one fiducial marker reduces 2D-3D spine image registration error slightly and improves robustness substantially. The findings are potentially relevant for image-guided therapy. If one marker is sufficient to obtain clinically acceptable registration accuracy and robustness, as the preliminary results using the proposed hybrid similarity measure suggest, the marker can be placed on a spinous process, which could be accomplished without penetrating muscle or using fluoroscopic guidance, and such a marker could be removed relatively easily.     

Automatic image quality quantification and mapping with an edge-preserving mask-filtering algorithm

Background: Imaging modalities in digital radiology produce large amounts of data for which image quality should be determined in order to validate the diagnostic operation.

Purpose: To develop an automatic method for image quality assessment.

Material and Methods: A filtering algorithm using a moving square mask was applied to create a map of filtered local intensity and noise values. Image quality scores were calculated from the filtered image data. The procedure was applied to technical and anthropomorphic (radiosurgery verification phantom [RSVP] head) phantom images obtained with varying radiation dose, field of view (FOV), and image content. The method was also applied to a clinical computed tomography (CT) brain image.

Results: The image quality score (IQs) of the phantom images increased from 0.51 to 0.82 as the radiation dose (CTDI_vol) increased from 9.2 to 74.3 mGy. Correlation of the IQs with the pixel noise was R² = 0.99. The deviation (1 SD) of IQs was 2.8% when the reconstruction FOV was set between 21 and 25 cm. The correlation of IQs with the pixel noise was R² = 0.98 with variable image contents and dose. Automatic tube current modulation applied to the RSVP phantom scan reduced the variation in the calculated image quality score by about 60% compared to the use of a fixed tube current.

Conclusion: The image quality score provides an efficient tool for automatic quantification of image quality. The presented method also produces a 2D image quality map, which can be used for further image analysis.     

Performance evaluation of an automated system for registration and postprocessing of CT scans

A prototype system optimized for automatic registration of CT scans and applications such as subtraction enhancement and CT angiography was evaluated. Co-registration was performed on 50 studies of various types, including CT angiograms. Mean registration speed was 109 s; accuracy was 2.0 mm. Examples of subtraction enhancement and automatic extraction of the vascular tree are provided. Unattended registration was demonstrated in routine application. Subtraction-enhanced images and extraction of the vascular tree may be useful in clinical practice.     

Diffusion magnetic resonance imaging patterns in metabolic and toxic brain disorders

Purpose: To evaluate metabolic and toxic brain disorders that manifest with restricted, elevated, or both restricted and elevated diffusion patterns on diffusion magnetic resonance imaging (MRI).

Material and Methods: Echo-planar diffusion MRI examinations were obtained in 34 pediatric patients with metabolic and toxic brain disorders proved by appropriate laboratory studies. The MRI unit operated at 1.5T with a gradient strength of 30 mT/meter, and a rise time of 600 μs. b=1000 s/mm² images and apparent diffusion coefficient (ADC) maps with ADC values were studied.

Results: Three patterns were observed: 1. A restricted diffusion pattern (high signal on b=1000 s/mm² images and low ADC values); 2. an elevated diffusion pattern (normal signal on b=1000 s/mm² images and high ADC values); and 3. a mixed pattern (coexistent restricted and increased diffusion patterns in the same patient). Disorders manifesting with a restricted diffusion pattern included metachromatic leukodystrophy (n=2), phenylketonuria (n=3), maple syrup urine disease (intermediate form) (n=1), infantile neuroaxonal dystrophy (n=1), Leigh (n=2), Wilson (n=3), and Canavan disease (n=1). Disorders with an elevated diffusion pattern included phenylketonuria (n=1), adrenoleukodystrophy (n=1), merosin-deficient congenital muscular dystrophy (n=2), mucopolysaccharidosis (n=2), Lowe syndrome (n=1), Leigh (n=2), Alexander (n=1), Pelizaeus-Merzbacher (n=1), and Wilson (n=3) disease. Disorders with a mixed pattern included L-2 hydroxyglutaric aciduria (n=2), non-ketotic hyperglycinemia (n=1), infantile neuroaxonal dystrophy (n=2), maple syrup urine disease (n=1), and Leigh (n=1) disease.

Conclusion: The findings suggested that the three different diffusion patterns reflect the histopathological changes associated with the disorders and different stages of a particular disorder. It is likely that the restricted diffusion pattern corresponds to abnormalities related to myelin, and the elevated diffusion pattern to disintegration of the tissue. The mixed pattern has contributions from both myelin abnormalities related to myelin disintegration of the tissue.     

Acquisition-related motion compensation for digital subtraction angiography

Subtraction methods in angiography are generally applied in order to enhance the visualization of blood vessels by eliminating bones and surrounding tissues from X-ray images. The main limitation of these methods is the sensitivity to patient movement, which leads to artifacts and reduces the clinical value of the subtraction images. In this paper we present a novel method for rigid motion compensation with primary application to road mapping, frequently used in image-guided interventions. Using the general concept of image-based registration, we optimize the physical position and orientation of the C-arm X-ray device, thought of as the rigid 3D transformation accounting for the patient movement. The registration is carried out using a hierarchical optimization strategy and a similarity measure based on the variance of intensity differences, which has been shown to be most suitable for fluoroscopic images. Performance evaluation demonstrated the capabilities of the proposed approach to compensate for potential intra-operative patient motion, being more resilient to the fundamental problems of pure image-based registration.