Endometrium estimation in a sequence of ultrasonic images.

In this paper a novel algorithm for motion estimation of the endometrium in a sequence of ultrasonic images is presented. The algorithm used is based on the criterion that a pixel of a motion object at different times has a different gray value. The motion estimation includes motion frequency and thickness; the former one is determined by the detection of change of a pixel value in a sequence of images during a given time period, and the latter is the result of the object's dimension divided by its extended width. The algorithm has been tested on synthetic and real images, and the results are encouraging.     

基于EM算法参数估计的各向异性扩散超声图像的去噪

目的 提出一种各向异性扩散滤波器的扩散参数选取方法,提高滤波器的灵活性和稳定性.方法 使用二状态的瑞利、高斯混合分布对超声图像灰度分布进行拟合,并采用期望值最大化(expectation maximization,EM)算法实现混合分布的分解;根据分解结果预测图像中斑点噪声均匀分布的区域;通过对均匀区域统计特性的分析获取各向异性扩散的扩散参数.结果 通过与两种改进扩散参数选取的滤波方法对比,基于EM算法的混合分布分解能够准确地估计扩散参数,使滤波结果在噪声消除和边缘保持上达到有效的平衡.结论 基于EM算法参数估计的各向异性扩散是一种有效的超声图像去噪方法.     

基于各向异性扩散的超声医学图像滤波方法

目的将各向异性扩散方法应用于超声医学图像的去噪处理。方法采用基于各向异性扩散的偏微分方程，其初始值为输入图像，转化为差分格式迭代求解滤波结果，并通过修改方程中的噪声比例函数使方程更加适用于超声医学图像的滤波。结果通过与其他3种传统滤波方法比较，修正系数后的各向异性扩散滤波方法能够较好地平滑图像的噪声，并且图像的边缘和细节部分依然清晰可见。结论各向异性扩散方法是一种去除超声图像Speckle噪声的有效方法，为超声医学图像的滤波开辟了一条新途径。     

Motion correction for myocardial T1 mapping using image registration with synthetic image estimation

Quantification of myocardial T1 relaxation has potential value in the diagnosis of both ischemic and nonischemic cardiomyopathies. Image acquisition using the modified Look-Locker inversion recovery technique is clinically feasible for T1 mapping. However, respiratory motion limits its applicability and degrades the accuracy of T1 estimation. The robust registration of acquired inversion recovery images is particularly challenging due to the large changes in image contrast, especially for those images acquired near the signal null point of the inversion recovery and other inversion times for which there is little tissue contrast. In this article, we propose a novel motion correction algorithm. This approach is based on estimating synthetic images presenting contrast changes similar to the acquired images. The estimation of synthetic images is formulated as a variational energy minimization problem. Validation on a consecutive patient data cohort shows that this strategy can perform robust nonrigid registration to align inversion recovery images experiencing significant motion and lead to suppression of motion induced artifacts in the T1 map.     

Reconstruction of coronary vessels from intravascular ultrasound image sequences based on compensation of the in-plane motion

A three-dimensional vessel model is reconstructed by fusing the cross-sectional information of vascular lumen detected from intravascular ultrasound (IVUS) frames with the spatial geometry of the ultrasonic catheter recovered from a pair of nearly orthogonal X-ray angiograms. This model is closer to the actual morphology than those reconstructed from angiograms or IVUS images alone because of the complementarity between angiography and IVUS in imaging coronary vessels. This study proposes a method to reconstruct the coronary vessels from an electrocardiogram (ECG)-gated IVUS image sequence and simultaneously acquired angiograms. The spatial orientation of each cross-section of the vascular lumen detected from IVUS frames was determined through quantitatively compensating the in-plane rigid motion caused by cardiac cycles. Independent validation of the determination of the IVUS spatial orientation with synthetic data was performed. A limited validity study including the back-projection validation and morphology measures with in vivo image data (five datasets) was performed to evaluate quantitatively the reconstruction accuracy.     

基于重构形态学算法的超声心脏图像自动分割

目的通过抑制超声医学图像本身固有的斑纹噪声,改进超声图像分割的算法,使得超声分割的结果更加准确。方法采用基于重构形态学的分割算法,先对超声目标图像进行重构开运算与闭运算,再对开闭结果进行top-hat运算,提取出相应亮度或暗度特征,最后找出亮暗度特征的边界实现全自动分割。结果利用本文方法对超声心脏图象进行自动分割,结果图象中没有因噪声而产生的伪边界,能够准确反映心脏各腔的真实情况。结论本文方法能够实现对超声心脏图象各腔边界的完整提取,分割边界具有很好的连续性。此外,本文方法也可以对目标的明或暗特征单独提取,从而减小时间复杂度,并提高分割图象的准确性。     

Frequency-domain simulation of MR tagging

Simulation of MR images is a useful tool for offline sequence development and as an aid to understanding image formation. One particular application of simulation is MR tagging, which is used for tracking myocardial motion. Simple spatial-domain methods cannot adequately represent effects common in these images, such as motion artifact and signal wrap. An existing frequency-domain model is shown to be inappropriate for tagged images, and an extension based on the Bloch equations and Fourier shift theorem is described to correct this. Software incorporating the new model is used to generate ideal tag intensity profiles and to accurately simulate tagged images. The shifted k-space patterns associated with tagged images, and their dependence on the order of the binomial tagging sequence, are explained. An application of the Fourier shift theorem is suggested that allows more rapid simulation of static tagged images.     

基于改进变分水平集法提取超声图像乳腺肿瘤边缘

目的研究准确提取超声图像乳腺肿瘤边缘的算法。方法在变分水平集方法中提出一种新型能量函数—全局—局部能量函数,通过加强图像的局部信息,实现超声图像乳腺肿瘤边缘的准确提取。结果对仿真超声图像的实验表明,本方法比全局最优化的Chan-Vese方法能更准确区分目标和背景。对103例超声乳腺肿瘤图像的实验表明,本方法无需任何预处理步骤即可准确提取肿瘤边缘。结论本方法有望实现超声图像乳腺肿瘤边缘的快速、准确提取。     

Artifacts due to stimulus correlated motion in functional imaging of the brain

To assess the effect of stimulus correlated motion on the appearance of functional magnetic resonance images, conventional visual and motor protocols were each performed by four normal volunteers and an image co-registration technique was used to retrospectively monitor subject motion. In three studies synthetic data sets were constructed from single baseline images using the positional information obtained from the co-registration procedure. Cumulative difference images were then created from both the synthetic and functional image sets. Stimulus correlated motion was detected in all eight studies and the synthetic cumulative difference images showed striking similarities to the equivalent functional images in each case.     

The effects of geometric distortion correction on motion realignment in fMRI

RATIONALE AND OBJECTIVES: Subject motion is well recognized as a significant impediment to resolution and sensitivity in functional magnetic resonance imaging (fMRI). A parallel confounder to fMRI data quality is geometric image distortion, particularly at high field strengths, due to susceptibility-induced magnetic field inhomogeneity. Consequently, many high-field echo-planar imaging methods incorporate a post-processing distortion correction by acquiring a field map of the sample prior to the fMRI measurement. However, field mapping methods impose a spatial mask on the data, since field information is only obtainable from regions with adequate signal-to-noise ratio (SNR). This masking, when applied to subsequent images in the fMRI time series, can clip the effects of motion, resulting in inaccurate estimation and correction of motion-based changes in the images. MATERIALS AND METHODS: The effects of geometric distortion correction on automated realignment (motion correction) of fMRI data are investigated from data acquired at 4 T. The results of image realignment with and without prior application of distortion correction are compared, using the estimated motion parameters and overall image realignment as metrics. RESULTS: The application of field-map-based distortion correction prior to image realignment reduces the amount of motion detected by a standard motion correction algorithm. Moreover, motion correction applied before distortion correction is shown to result in superior realignment of motion-correction images. CONCLUSION: It is preferable to perform motion realignment prior to correcting for geometric distortion.     

A self-learning segmentation framework--the Taguchi approach.

The detection of object boundary is an interesting and challenging task in computer vision and medical image processing. The active contour model (snake model) has attracted much attention for object boundary detection in the past decade. However, due to the lack of understanding on the effect of different energy terms to the behavior of related objective functions for an image, the assignment of weights for different energy terms in this model is usually fulfilled empirically. Few discussions have been brought out specifically for assigning these weights automatically. In this paper, a novel self-learning segmentation framework, based on the snake model is proposed and applied to the detection of cardiac boundaries from ultrasonic images. The framework consists of a learning section and a detection section, and provides a training mechanism to obtain the weights from a desired object contour given manually. This mechanism first employs Taguchi's method to determine the weight ratios among distinct energy terms, followed by a weight refinement step with a genetic algorithm. The refined weights can be treated as the a priori knowledge embedded in the manually defined contour and be used for subsequent contour detection. Experiments with both synthetic and real echocardiac images were conducted with satisfactory outcomes. Results also show that the present method can be used to analyze successive images of the same object with only one training contour. Finally, the validity of the weight determining process was verified by the analysis of variance method (ANOVA).     

Adaptive-weighted bilateral filtering and other pre-processing techniques for optical coherence tomography

This paper presents novel pre-processing image enhancement algorithms for retinal optical coherence tomography (OCT). These images contain a large amount of speckle causing them to be grainy and of very low contrast. To make these images valuable for clinical interpretation, we propose a novel method to remove speckle, while preserving useful information contained in each retinal layer. The process starts with multi-scale despeckling based on a dual-tree complex wavelet transform (DT-CWT). We further enhance the OCT image through a smoothing process that uses a novel adaptive-weighted bilateral filter (AWBF). This offers the desirable property of preserving texture within the OCT image layers. The enhanced OCT image is then segmented to extract inner retinal layers that contain useful information for eye research. Our layer segmentation technique is also performed in the DT-CWT domain. Finally we describe an OCT/fundus image registration algorithm which is helpful when two modalities are used together for diagnosis and for information fusion.     

Turboprop IDEAL: A motion‐resistant fat–water separation technique

Suppression of the fat signal in MRI is very important for many clinical applications. Multi‐point water–fat separation methods, such as IDEAL (Iterative Decomposition of water and fat with Echo Asymmetry and Least‐squares estimation), can robustly separate water and fat signal, but inevitably increase scan time, making separated images more easily affected by patient motions. PROPELLER (Periodically Rotated Overlapping ParallEL Lines with Enhanced Reconstruction) and Turboprop techniques offer an effective approach to correct for motion artifacts. By combining these techniques together, we demonstrate that the new TP‐IDEAL method can provide reliable water–fat separation with robust motion correction. The Turboprop sequence was modified to acquire source images, and motion correction algorithms were adjusted to assure the registration between different echo images. Theoretical calculations were performed to predict the optimal shift and spacing of the gradient echoes. Phantom images were acquired, and results were compared with regular FSE‐IDEAL. Both T1‐ and T2‐weighted images of the human brain were used to demonstrate the effectiveness of motion correction. TP‐IDEAL images were also acquired for pelvis, knee, and foot, showing great potential of this technique for general clinical applications. Magn Reson Med 61:188–195, 2009. © 2008 Wiley‐Liss, Inc.     

Reconstruction of undersampled dynamic images by modeling the motion of object elements.

Dynamic MRI is restricted due to the time required to obtain enough data to reconstruct the image sequence. Several undersampled reconstruction techniques have been proposed to reduce the acquisition time. In most of these techniques the nonacquired data are recovered by modeling the temporal information as varying pixel intensities represented in time or in temporal frequencies. Here we propose a new approach that recovers the missing data through a motion estimation of the object elements ("obels," or pieces of tissue) of the image. This method assumes that an obel displacement through the sequence has lower bandwidth than fluctuations in pixel intensities caused by the motion, and thus it can be modeled with fewer parameters. Preliminary results show that this technique can effectively reconstruct (with root mean square (RMS) errors below 4%) cardiac images and joints with undersampling factors of 8 and 4, respectively. Moreover, in the reconstruction process an approximation of the motion vectors is obtained for each obel, which can be used to quantify dynamic information. In this method the motion need not be confined to a part of the field of view (FOV) or to a portion of the temporal frequency. It is appropriate for dynamic studies in which the obels' motion model has fewer parameters than the number of acquired samples.     

Three-dimensional reconstruction of fine vascularity in ultrasound breast imaging using contrast-enhanced spatial compounding: in vitro analyses

RATIONALE AND OBJECTIVES: Ultrasound image quality can be improved by imaging an object (here: the female breast) from different viewing angles in one image plane. With this technique, which is commonly referred to as spatial compounding, a more isotropic resolution is achieved while speckle noise and further artifacts are reduced. We present results obtained from a combination of spatial compounding with contrast-enhanced ultrasound imaging in three dimensions to reduce contrast specific artifacts (depth dependency, shadowing, speckle) and reconstruct vascular structures. MATERIALS AND METHODS: We used a conventional ultrasound scanner and a custom made mechanical system to rotate an ultrasound curved array probe around an object (360 degrees , 36 transducer positions). For 10 parallel image planes, ultrasound compound images were generated of a flow-mimicking phantom consecutively supplied with water and contrast agent. These compound images were combined to form a volume dataset and postprocessed to obtain a sonographic subtraction angiography. RESULTS: Image quality was significantly improved by spatial compounding for the native (ie, without contrast agent), and, in particular, for the contrast-enhanced case. After subtracting the native images from the contrast-enhanced ones, only structures supplied with contrast agent remain. This technique yields much better results for compound images than for conventional ultrasound images because speckle noise and an anisotropic resolution affect the latter. CONCLUSIONS: With the presented approach contrast specific artifacts can be eliminated efficiently, and a subtraction angiography can be computed. A speckle reduced three-dimensional reconstruction of submillimeter vessel structures was achieved for the first time. In the future, this technique can be applied in vivo to image the vascularity of cancer in the female breast.     

Advantages of real-time spatial compound sonography of the musculoskeletal system versus conventional sonography

OBJECTIVE: Spatial compound sonography is a method that obtains sonographic information from several different angles of insonation and combines them to produce a single image. By reducing speckle and improving definition of tissue planes, this method can potentially improve image quality in musculoskeletal sonography. The purpose of our study was to compare real-time spatial compound sonography with conventional high-resolution musculoskeletal sonography. MATERIALS AND METHODS: Thirty-four patients underwent sonography of the musculoskeletal system for a variety of indications. All patients were evaluated using conventional high-resolution sonography and real-time spatial compound sonography performed with a 12-5-MHz multifrequency linear array transducer. Conventional images and compound images depicting the same musculoskeletal structure were obtained in pairs. A total of 118 images (59 image pairs) were randomly assorted and reviewed on a computer monitor by three experienced sonologists working independently. The reviewers were unaware of the type of images they were evaluating. Image quality was rated using a 5-point scale. The image parameters evaluated were definition of tissue planes, speckle, other noise, and image detail. RESULTS: Analysis of variance revealed that real-time spatial compound sonography significantly improved definition of soft-tissue planes, reduced speckle and other noise, and improved image detail when compared with conventional high-resolution sonography (p < 0.0001 for all evaluated parameters). CONCLUSION: Real-time spatial compound sonography significantly improved sonographic image quality in the musculoskeletal system when compared with conventional high-resolution sonography. Because musculoskeletal sonography is highly dependent on image quality and tissue-plane definition, spatial compound sonography represents an important development.     

Real-time bronchoscope three-dimensional motion estimation using multiple sensor-driven alignment of CT images and electromagnetic measurements

Bronchoscope three-dimensional motion estimation plays a key role in developing bronchoscopic navigation systems. Currently external tracking devices, particularly electromagnetic trackers with electromagnetic sensors, are increasingly introduced to navigate surgical tools in pre-clinical images. An unavoidable problem, which is to align the electromagnetic tracker to pre-clinical images, must be solved before navigation. This paper proposes a multiple sensor-driven registration method to establish this alignment without using any anatomical fiducials. Although current fiducially free registration methods work well, they limit to the initialization of optimization and manipulating the bronchoscope along the bronchial centerlines, which could be failed easily during clinical interventions. To address these limitations, we utilize measurements of multiple electromagnetic sensors to calculate bronchoscope geometric center positions that are usually closer to the bronchial centerlines than the sensor itself measured positions. We validated our method on a bronchial phantom. The experimental results demonstrate that our idea of using multiple sensors to determine bronchoscope geometric center positions for fiducial-free registration was very effective. Compared to currently available methods in bronchoscope three-dimensional motion estimation, our method reduced fiducial alignment error from at least 6.79 to 4.68–5.26 mm and significantly improved motion estimation or tracking accuracy from at least 5.42 to 3.78–4.53 mm.     

MR imaging of lung parenchyma: a solution to susceptibility.

The authors have developed a pulse sequence for imaging lung parenchyma with projection reconstruction magnetic resonance (MR) imaging that reduces the effects of motion and susceptibility. In this study, the projection reconstruction technique was further modified by optimizing MR signal frequencies for reconstructing the images. This was done by means of one of two methods. With the first method, a susceptibility map was derived from the raw image data and this map was used to indicate the optimal frequencies for reconstructing the images. The second method of susceptibility correction was a postprocessing technique in which the optimal reconstruction frequencies were selected with use of specific focusing criteria to generate the least blurred image. The effect of using susceptibility map correction on a phantom was demonstrated, and both of these methods were used to improve the visibility of pulmonary structures on images of subjects with normal and abnormal lungs.     

The contrast-to-noise in relaxation time, synthetic, and weighted-sum MR images

The contrast-to-noise ratio (CNR) in three types of computed MR images is compared in a computer simulation. The original data consist of two spin-echo or two saturation-recovery images. Each pair of images is used to generate a relaxation time image, a synthetic image at arbitrary echo or repetition time, and an image which is a weighted sum of the original images. The CNR produced by these three methods is compared for signals spanning a wide range of relaxation times. In every comparison an optimally weighted sum produces the highest CNR that is statistically attainable. The CNR in the optimum synthetic image equals this bound only if contrast reversal does not occur in the original images. The CNR in relaxation time images is always less than the statistical bound and can be less than the CNR in the original images.