【作者】 金朝阳；

【导师】 夏灵； 杜一平；

【作者基本信息】 浙江大学 ， 生物医学工程， 2011， 博士

【摘要】 近年来研究表明,磁敏感加权成像(Susceptibility Weighted Imaging, SWI)对颅内静脉损伤和疾病的诊断具有非常重要的临床价值。SWI是一种三维、高分辨率、T2*加权的梯度回波成像方式。它利用静脉和组织之间磁敏感性差异引发的相对相位变化和幅值改变来增强小静脉的可见性。为了获得接近最优的静脉对比,SWI常选择较长的回波时间,因而数据采集时间较长。回波时间长时,会引起T2*衰减,导致图像的信噪比相对较低。由于梯度回波对磁场不均匀性比较敏感,因此在构建三维SWI相位模板时,某些局部磁场严重不均匀的区域会因为非共振效应而产生伪影。由于静脉负对比度,在三维SWI数据显示时常采用最小密度投影(Minimum-Intensity Projection, mIP),在轴向投影中,由于大脑的自然形状,低密度的空气和骨组织体素会出现在脑外围区域的投影线中,导致mIP图像中脑外围区域的静脉和脑组织信号损失而产生伪影。低密度的空气和骨组织也使得常规全脑SWl数据的矢状面和冠状面投影难以进行。此外,小静脉的可见性与背景信号密切相关,抑制背景信号能增强小静脉的可见性。本文针对上述问题,对SWI快速数据采集和重建、伪影校正和静脉增强等问题做了较为系统和深入的研究,主要包括：(1)研究了部分k-空间数据采集和基于凸集投影的图像重建方法,能有效减少数据采集时间,并表明在部分k-空间中施加三维Fermi滤波器,可提高图像的信噪比和减小空间分辨率的角度相关性。(2)提出了双回波和多回波磁共振动静脉数据同时采集(Simultaneous Acqusition of Magnetic Resonance Angiography and Venography, MRAV)技术,在3T下采集MRA(Magnetic Resonance Angiography)数据的同时,无需增加额外的扫描时间,就可获得MRV (Magnetic Resonance Venography)数据。研究了空间分辨率对MRV图像静脉-背景对比度的影响。指出不同长度回波时间(Echo Time, TE)对MRV图像信噪比和非共振伪影的影响。表明通过拟合多个TE数据能更有效地映射R2*信号,进而量化脑区的铁离子沉积度。(3)研究了一种基于局部磁场梯度计算的非共振伪影校正方法,无需进行相位解卷绕,通过计算局部磁场梯度来估计局部磁场的不均匀程度,进而抑制相位模板中残留的背景信息。(4)研究了两类脑组织体分割方法：一种是基于局部相位和幅值统计属性构建多变量映射来提高空气和脑组织的分离度,从而更可靠地分割脑组织；另一种是基于改进的变分水平集算法,直接对幅值图像中的脑组织进行分割。这两种方法都能有效地抑制三维SWI数据越面显示时脑外围区域的信号损失和进行面内mIP。(5)研究了三种小静脉图像增强算法：构建了图像域的二阶相位差高通滤波器来增强小静脉的相移,并用Fermi滤波器抑制SWI的背景信号；研究了对幅值图像的k-空间数据进行高通滤波来抑制背景组织的方法,提高小静脉的可见度；提出了基于Hessian矩阵的三维多尺度静脉增强算法,通过研究Hessian矩阵的特征值对静脉的特异性表现,来抑制背景组织和噪声,增强小静脉的可见性。更多还原

【Abstract】 Susceptibility-weighted imaging (SWI) has recently demonstrated great clinical significance in the diagnosis of several intracranial venous lesions and diseases. SWI is a 3D, high resolution, T2*-weighted, and gradient-echo imaging technique. SWI utilizes the relative phase and magnitude change in the venous vasculature introduced by the susceptibility difference between venous blood and parenchyma. A relatively long echo time (TE) is typically used in SWI to achieve nearly optimal venous contrast, resulting in long data acqusition time. The SWI data have relatively low signal-to-noise ratio (SNR) due to the T2* decay at a long TE. Artifact can arise during the construction of the 3D phase mask in regions with severe field inhomogeneity due to the off-resonance effect. Image artifacts can also arise from minimum-intensity projection (mIP), which is commonly used for the display of venous vasculature because of the negative venous contrast. In axial projection, voxels in air and bone can be in the path of projection in peripheral regions of the brain due to the nature shape of the brain. The low intensity in air or bone results in the disappearance of signal from brain tissue, including veins, in mIP images in that region. The low intensity of air and bone also make mIP impractical for sagittal and coronal projections of whole-brain SWI data. Moreover, visualization of the venous vasculature in the magnitude data can be enhanced by background suppression. In this study, we investigated following aspects:(1) The partial k-space acquisition was studied to reduce scan time and the magnitude images were reconstructed by the projection onto convex sets algorithm. A 3D Fermi filter in k-space was demonstrated to increase the SNR and to reduce angular dependence of spatial resolution.(2) A dual-echo and multi-echo pulse sequence for simultaneous acquisition of magnetic resonance angiography and venography (MRAV) were developed. Using this pulse sequence, the magnetic resonance venography (MRV) data can be acquired without increasing the scan time of magnetic resonance angiography (MRA). The effect of spatial resolution on vein-to-background contrast was demonstrated. The venous contrast and off-resonance artifacts of MRV data acquired at different TEs were studied. The R2* value at each voxel was quantified using multi-TE exponential fitting. The R2* map can be used for the quantification of iron deposition.(3) A novel postprocessing approach was studied to calculate the local field gradient (LFG) for the reduction of the off-resonance artifacts without phase-unwrapping. LFG measurements were used to assess the severity of field inhomogeneity and suppress the residual phase in the phase mask induced by the off-resonance effect.(4) Two volume segmentation algorithms of the brain tissue were studied. A multivariate measure based on the statistics of phase and magnitude was constructed for robust tissue-air volume segmentation. An improved version of the variational level set algorithm was used for volume segmentation of brain tissue directly based on magnitude images. Both algorithms provide a feasible solution to reduce the signal loss in the peripheral regions of the brain in the through-plane mIP images and enable in-plane mIP display of MRV.(5) Three enhancement algorithms of small vein were presented. First, the image-domain high-pass filters based on second-order phase difference were applied to the complex 3D SWI data to enhance the susceptibility phase shift of the veins and suppress background signal in SWI. Second, high-pass filter was applied to the Fourier domain of the magnitude images to suppress the background signal and enahnce the visibility of the venous vasculature in the brain. Third, a 3D multi-scale vessel enhancement algorithm based on the Hessian matrix was proposed. Eigenvalue analysis of the Hessian matrix was used to enhance the veins, suppress the background tissue, and reduce the noise in air.更多还原