DOI: 10.1148/rg.283075715
RadioGraphics 2008;28:639-651
© RSNA, 2008
Informatics in Radiology
GUIBOLD: A Graphical User Interface for Image Reconstruction and Data Analysis in Susceptibility-weighted MR Imaging1
Andreas Deistung, MSc,
Alexander Rauscher, PhD,
Jan Sedlacik, PhD,
Stephan Witoszynskyj, MSc, and
Jurgen R. Reichenbach, PhD
1 From the Medical Physics Group, Institute of Diagnostic and Interventional Radiology (A.D., A.R., J.S., S.W., J.R.R.), and the Core Unit MR Methods (A.D., A.R., J.S.), University Hospital Jena, Friedrich-Schiller-University, MRI-Building "Am Steiger," Philosophenweg 3, D-07743 Jena, Germany; and the MRI Research Centre, University of British Columbia, Vancouver, British Columbia, Canada (A.R.). Received April 26, 2007; revision requested August 29 and received October 5; accepted January 9, 2008. Supported by grant RE 1123/7-1 and 7-2 from the Deutsche Forschungsgemeinschaft, by the Interdisciplinary Center for Clinical Research, and by grants COST-STSM-B21-01283 and COST-STSM-B21- 00305 from the COST B21 action Short Term Scientific Missions. All authors have no financial relationships to disclose.
Address correspondence to A.D. (e-mail: andreas.deistung{at}med.uni-jena.de).
 |
Abstract
|
|---|
Susceptibility-weighted (SW) magnetic resonance (MR) imaging provides high-resolution, distortion-free blood oxygen level–dependent (BOLD) data for assessment of cerebral veins, blood products, and brain lesions. Currently, reconstruction of SW imaging data is not implemented on all MR imaging systems or is restricted in terms of parameter adjustments. New developments in SW imaging have been implemented into a graphical user interface (GUI), which is named GUIBOLD. The GUI was designed for imaging system–independent off-line data reconstruction with interactive setting of parameters on the basis of k-space data and Digital Imaging and Communications in Medicine images. GUIBOLD is capable of presenting magnitude, unwrapped phase, and SW images in different orientations and parallel projections with various rendering methods and region-of-interest–based data analysis tools. Moreover, GUIBOLD affords easy and comprehensive data reconstruction possibilities for venographic and arterial imaging and anatomic phase imaging. As a direct application, differentiation between cavernous and calcified lesions on the basis of their magnetic susceptibility on phase images was performed. GUIBOLD widens the range of potential applications of SW imaging and makes it more accessible for use in the clinical routine as well as in medical research.
© RSNA, 2008
|
Introduction
|
|---|
Susceptibility-weighted (SW) imaging (1) provides high-resolution and distortion-free images that are sensitive to the local tissue magnetic susceptibility and the blood oxygen level–dependent (BOLD) effect (2). Although the original main application of SW imaging was to visualize cerebral veins at the submillimeter level without application of contrast agents (3), it was quickly realized that the inherent sensitivity of SW imaging to deoxygenated blood and other susceptibility differences allows further applications as well. These include detection of cerebral hemorrhages (4) and stroke (5), localization and classification of tumors (6–9), imaging of calcium deposits or iron products, such as ferritin and hemosiderin (9,10), as well as retrieval of functional information from modifications of blood oxygenation by breathing carbogen (11) or by ingestion of caffeine (12).
SW imaging takes advantage of the dependence of gradient-echo magnetic resonance (MR) signals on venous blood oxygenation (the BOLD effect). Because venous (deoxygenated) blood is less diamagnetic than arterial (oxygenated) blood, veins produce local field inhomogeneities in their surroundings, which cause spins to dephase faster than in a homogeneous field. Ultimately, this effect produces local signal reduction. In addition to dephasing, structures with different magnetic susceptibility also produce offsets in the resonance frequency, which are reflected in the phase images. Consequently, visibility of tissue or vascular structures with different magnetic susceptibility with respect to the surroundings in which they are embedded can be dramatically improved if both magnitude information and phase information are combined to create SW images.
SW imaging reconstruction and data visualization involve a number of postprocessing methods, including phase correction, computation of a phase mask, and the combination of phase and magnitude, which are usually not (or not freely) available on routine imaging systems. So far, no intuitive, fast, and versatile off-line software exists for processing SW imaging data, to our knowledge. On the MR systems on which SW imaging acquisition and data reconstruction are currently available, the user has only limited control in the choice of reconstruction parameters. Furthermore, with fixed data processing software, it becomes more difficult for MR system manufacturers to react quickly to new developments, such as improvements of or alternative approaches to phase unwrapping (13), median filtered SW imaging (14), or special applications of phase imaging (15).
In an effort to overcome these limitations, we designed a flexible graphical user interface (GUI), herein referred to as GUIBOLD, for SW image reconstruction, advanced data processing, and visualization. We demonstrate its use in a number of clinical applications and present its application to a novel technique that allows lesion classification on the basis of the magnetic susceptibility of a lesion. The software is continuously updated and offers full control over all image reconstruction and postprocessing parameters.
|
Acquisition of SW Data
|
|---|
To illustrate the use of GUIBOLD, SW data from the human brain were acquired with a high-resolution, first-order fully flow-compensated, T2*-weighted three-dimensional (3D) sequence on both a 1.5-T and 3-T MR imaging system. High spatial resolution is required to increase the partial volume effect produced by small veins (16) and to ensure that macroscopic field inhomogeneities with low spatial frequencies do not lead to intravoxel dephasing and hence to unwanted signal loss (17). Inflow effects of blood are effectively reduced by exciting thick 3D slabs as well as by flow compensation in all three directions. The latter also ensures that stationary and moving spins refocus at the same time.
Because local phase offsets associated with differences in magnetic susceptibility increase with echo time (TE), relatively long TEs are required with the sequence to be sensitive to small local field inhomogeneities (eg, those produced by venous vessels). Typical imaging parameters at 1.5 T are as follows: repetition time (TR) = 57–67 msec, TE = 40 msec, flip angle = 20°–25°, field of view = 192 x 256 x 64 mm, matrix = 256 x 512, and section thickness = 1.5–2 mm. There are two possibilities to reduce the long TR while still preserving the same venous vessel contrast. One possibility is the administration of paramagnetic contrast agents to patients to shorten the TE and thus the TR (18,19). The other opportunity is to use the field dependence of signal cancellation effects due to susceptibility differences, which again enables a shorter TE with increasing field strength and results in the following typical acquisition parameters at 3 T: TR/TE = 35 msec/25 msec, flip angle = 15°.
|
Software Design
|
|---|
GUIBOLD was developed under the Linux operating system and written in the IDL programming language (Interactive Data Language, version 6 and 6.2; ITT Visual Information Solutions, Boulder, Colo). It was designed for an easy offline reconstruction of SW images on conventional personal computer hardware. IDL integrates a powerful, array-oriented language with numerous mathematical analysis and graphical display techniques (20). The source code of GUIBOLD was then compiled for the freely available IDL virtual machine. The IDL virtual machine makes it possible to run GUIBOLD as a cross-platform application (Windows [Microsoft, Redmond, Wash], Linux, and Mac OS X [Apple Computer, Cupertino, Calif]) without having to purchase an IDL license. The advantages of an off-line reconstruction include independence of a particular MR imaging system and multiple possibilities to reconstruct measured data with different postprocessing settings, which can be easily adapted to the special needs of the user. This makes SW imaging accessible to basic clinical research and routine, where straightforward yet highly flexible data processing is essential.
Figure 1 is a flowchart of the user interface’s data processing methods. The first step in the handling of GUIBOLD is import of data. GUIBOLD accepts both raw k-space data and DICOM images. To ensure a straightforward data processing pipeline behind the user interface, DICOM data are reconverted via Fourier transform to k-space data. The algorithms used in GUIBOLD for SW image reconstruction can be divided into two major parts: k-space data preprocessing and SW image computation.
View larger version (24K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1. Flowchart of the GUI-integrated algorithms for SW imaging data processing. DICOM = Digital Imaging and Communications in Medicine, FFT = fast Fourier transform, 2D = two-dimensional.
|
|
k-Space Data Preprocessing
The Reconstruction Window (Fig 2a) represents the central user interface, which allows access to all parameters of the employed reconstruction steps and algorithms, such as zero filling, windowing Fourier transform, and phase unwrapping by using homodyne filtering or region growing. Zero filling is applied to k-space data to obtain isotropic in-plane resolution of the image data. Subsequent multiplication of k-space data with a Kaiser-Bessel window (21) suppresses Gibbs ringing.
View larger version (38K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2a. User interface of GUIBOLD. (a) The Reconstruction Window allows interactive setting of parameters for SW image reconstruction. (b) Computer screen displays an axial venogram (minimum intensity projection [mIP] obtained over 14 mm), which shows multiple cavernomas (dark spots close to vessels). The left side of the window shows the Image Display Menu, where the visualization options for the reconstructed data are chosen. First, the image type (magnitude, phase, or venogram) is selected; subsequently, the user has access to different imaging sections with a slider. Projection and volume imaging can then be selected. In addition, windowing and zooming as well as region of interest–based analysis are implemented.
|
|
View larger version (99K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2b. User interface of GUIBOLD. (a) The Reconstruction Window allows interactive setting of parameters for SW image reconstruction. (b) Computer screen displays an axial venogram (minimum intensity projection [mIP] obtained over 14 mm), which shows multiple cavernomas (dark spots close to vessels). The left side of the window shows the Image Display Menu, where the visualization options for the reconstructed data are chosen. First, the image type (magnitude, phase, or venogram) is selected; subsequently, the user has access to different imaging sections with a slider. Projection and volume imaging can then be selected. In addition, windowing and zooming as well as region of interest–based analysis are implemented.
|
|
The Fourier transform of k-space data produces a complex MR image consisting of magnitude and phase. Because the phase is defined only in the range of [–
, ), phase values that differ by multiples of 2 cannot be distinguished. This leads to phase wraps in the phase images, which require a correction—a process commonly referred to as phase unwrapping.
GUIBOLD provides three different phase unwrapping methods: homodyne filtering (22) in 2D (left thread in Fig 1) and 3D (right thread in Fig 1) as well as phase unwrapping by region growing (
UN [23]) (13,24,25) (middle thread in Fig 1). The method of phase unwrapping is selected in the bottom row in the Reconstruction Window (Fig 2a). If no phase unwrapping method is selected, wrapped phase images are used to compute the final SW image.
Homodyne filtering is a complex division in image space of the original complex data by the same low-pass filtered data. In GUIBOLD, homodyne 3D filtering (right thread) is implemented based on low-pass filtered data obtained by multiplication of k-space with a 3D Hanning function.
For homodyne filtering in 2D (left thread), a Fourier transform is applied first along the section partitioning direction. Next, each section is multiplied by 2D Hanning functions followed by section-by-section 2D Fourier transform, resulting in low-pass filtered complex images.
The region-growing algorithm (middle thread) (23,25) unwraps the 2D phase image on a pixel-by-pixel base by computing linear predictions from already unwrapped pixels. Low-frequency phase variations caused by static magnetic field inhomogeneities are then removed by applying a 2D high-pass filter (13). In the resulting phase image, effects from small structures with different magnetic susceptibility are preserved, whereas phase variations due to background field inhomogeneities are suppressed.
Homodyne filtering is an established and robust method. In GUIBOLD, the 2D version is set as the default due to its small number of parameters and its lower vulnerability to strong field inhomogeneities compared to the 3D homodyne technique.
Region-growing phase unwrapping is an alternative that improves phase correction, particularly near regions with large susceptibility jumps, such as air-tissue or bone-tissue interfaces. Another advantage of region growing is that it preserves all spatial frequencies of phase information, whereas homodyne filtering always produces a high-pass filtered phase image. This adds flexibility to further processing steps. For instance, with subsequent high-pass filtering, a parameterizable shifting of the cutoff spatial frequencies is possible and allows differentiation between low-frequency structures (different tissue types) and high-frequency structures like small venous vessels.
Computation of SW Images
Magnitude and unwrapped phase images are required to compute SW images. The conventional phase mask used in SW imaging is obtained from unwrapped phase images, weighted with a function that scales negative phase values linearly between 0 and 1 and sets positive values, representing arteries and parenchyma, to unity (17). A triangular phase mask function is also implemented in GUIBOLD to improve vessel visualization in coronal or sagittal SW imaging volumes. The final SW image is then obtained by performing a user-defined number of multiplications of the phase mask with the corresponding magnitude image. Fourfold multiplication is the default setting (26). Phase masking and its combination with magnitude data can be adjusted in the SW image computing section of the Reconstruction Window (Fig 2a).
Data Visualization
After data processing, the Image Display Menu appears on the left side of the GUI (Fig 2b). Here, the user can change between different contrasts of magnitude, phase, and SW images.
These images are displayed in GUIBOLD section by section as well as in different conventional parallel projections (maximum intensity projection [MIP], average projection, and minimum intensity projection [mIP]) over adjustable volumes in axial, sagittal, and coronal orientations. The projections are computed by selecting or calculating the maximum (MIP), average, or minimum (mIP) intensity along orthogonal parallel rays traversing the volume. The MIP procedure is also implemented with freely selectable projection angles for magnitude, phase mask, and SW images.
Figure 2b shows a venogram from a patient with multiple cavernous lesions. In this case, an mIP over 14 mm of the axial SW imaging data set was created via the Image Display Menu. The user can scroll through the data by using a slider. The IDL volume visualization tools XVOLUME (27) and SLICER3 (28) can be called from the Image Display Menu. With both of these tools, GUIBOLD facilitates additional methods for viewing 3D data sets as well as manipulating volumes and isosurfaces interactively.
Data Evaluation
Via the "Measurements" button in the menu bar (Fig 2b), quantitative data evaluation is possible by region-of-interest–based analysis with multiple regions of interest of different shapes (eg, rectangle, square, arbitrary). The region-of-interest location can be written to and read from files. This facilitates analysis of all modalities (magnitude, phase, and SW data). The results of the analysis can also be saved. Distance and angle measurements are integrated to characterize and analyze vessels by their length or the angles of branches.
Data Requirements
GUIBOLD supports many different data formats, including complex k-space data files with and without a second, independent header file in Analyze format (29) and Siemens Medical Solutions (Erlangen, Germany) raw data objects (currently for the imaging system software versions up to VB12 and VA25). The latter consist of a data file (meas.out) and a corresponding information file (meas.asc). At our institution, all SW imaging sequences running on the Magnetom Vision, Magnetom Sonata, and TIM Trio MR systems (Siemens Medical Solutions) are modified to produce k-space data output that is readable with GUIBOLD. Furthermore, system-independent DICOM images (in image space) with magnitude and phase information (wrapped or unwrapped after phase correction implemented on the particular MR system) are also supported.
Finally, images reconstructed with GUIBOLD can be saved in either DICOM or Analyze format (29). Because DICOM is supported, long-term storage in picture archiving and communication systems is possible. GUIBOLD also supports output in several common image formats (PNG, JPEG, BMP, TIFF, GIF, PPM, SRF).
|
Clinical Applications
|
|---|
Venous Imaging
Evaluation and visualization of SW data with GUIBOLD allows one to differentiate between the arterial vessel architecture (as seen on the magnitude images) and the venous vessel architecture (as seen on SW images) (Fig 3). Often this differentiation is clinically useful because verification of the presence or absence of venous blood can be important for the evaluation of lesions (1). The appraisal of venous vessels also plays a fundamental role in the observation of tumor vascularity (6,7) and diagnosis of arteriovenous malformations (AVMs) (30) or cerebral venous thromboses (31). For instance, cortical venous thromboses appear hypointense on T2*-weighted images (32) from onset to the 7th day of the thrombus. However, the thrombus may be less pronounced or even not visible during stages in which no paramagnetic substances are present or the compartmentalization of magnetic susceptibility gradients is reduced (33).
View larger version (52K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3. GUIBOLD enables MIP display of magnitude images (left) and mIP display of SW images (right). Therefore, GUIBOLD allows simultaneous visualization of arterial and venous vessels in corresponding volumes. This visualization mode allows one to perform a simple artery-vein separation owing to the presence of two distinct contrast mechanisms.
|
|
Venous vessels are mapped as hypointense structures on SW images. mIPs over typically 5–20 mm of SW data are usually computed to create MR-BOLD venograms. Figure 2b demonstrates the visualization of numerous cerebral veins via an mIP of SW data with GUIBOLD. This venogram also highlights multiple cavernomas, which are filled with deoxygenated blood. Owing to the 3D SW imaging sequence scheme, single sections as well as projections in all three directions can be displayed. The default setting is that GUIBOLD weights the axial images with the conventional phase mask and the two other orientations with triangular phase masks (Fig 4) because of the angle-dependent field distribution of venous structures with respect to the static magnetic field (1).
View larger version (132K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4a. SW images of a volunteer. The images were projected over 17 mm and weighted with a normal phase mask in the axial orientation (a) and with a triangular phase mask in the coronal (b) and sagittal (c) orientations. The white markers in a indicate the corresponding projection volumes for b and c.
|
|
View larger version (94K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4b. SW images of a volunteer. The images were projected over 17 mm and weighted with a normal phase mask in the axial orientation (a) and with a triangular phase mask in the coronal (b) and sagittal (c) orientations. The white markers in a indicate the corresponding projection volumes for b and c.
|
|
View larger version (82K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4c. SW images of a volunteer. The images were projected over 17 mm and weighted with a normal phase mask in the axial orientation (a) and with a triangular phase mask in the coronal (b) and sagittal (c) orientations. The white markers in a indicate the corresponding projection volumes for b and c.
|
|
If structures of interest are restricted predominantly to venous vessels, GUIBOLD offers two possibilities. First, median filtered SW imaging data (14,34) can be processed and displayed in mIPs (Fig 5). In this case, the SW imaging volume is nonlinearly filtered with a median kernel in 3D and then subtracted from the original SW imaging data set to improve the in-plane visualization of venous vessels by suppressing low-frequency tissue structures. Second, GUIBOLD offers MIPs of preprocessed SW imaging data (Fig 6a). These preprocessing steps include brain extraction from the skull with a threshold-based region-growing procedure, followed by inversion of data to enable the MIP procedure with freely adjustable projection angles through the whole brain.
View larger version (184K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5. mIP image obtained over 15 mm of median filtered SW data (14). DICOM data were imported from a 3-T system and conventional SW images were reconstructed, followed by subtraction of median filtered SW imaging data from the original SW imaging data to enhance high-frequency vascular structures. Finally, mIP of the subtraction data was performed. Note that even small penetrating venous vessels in subcortical regions are visible.
|
|
View larger version (101K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6a. (a) On an MIP image from masked and inverted SW imaging data, venous vessels are highlighted. Arrowheads show regions where the masking algorithm failed. (b) On an MIP image from magnitude data, mainly arteries are visible. The sagittal sinus is highlighted due to the high venous outflow.
|
|
View larger version (72K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6b. (a) On an MIP image from masked and inverted SW imaging data, venous vessels are highlighted. Arrowheads show regions where the masking algorithm failed. (b) On an MIP image from magnitude data, mainly arteries are visible. The sagittal sinus is highlighted due to the high venous outflow.
|
|
Arterial Imaging
Magnitude information and contrast are influenced by intrinsic tissue parameters, such as proton density, T1, and T2 or alternatively T2*. On T2*-weighted magnitude images, arteries appear as high-signal-intensity structures due to their longer T2* values compared with those of parenchyma and venous blood. Thus, these vessels can be highlighted by GUIBOLD on the basis of MIPs (arteriograms) to provide the radiologist with additional information about the vasculature (Fig 6b).
The simultaneous display of corresponding arteriograms and venograms (Fig 3) makes it possible to assess feeding and draining vessels to the brain. Although these arteriograms certainly cannot compete in image quality with dedicated, specifically tailored time-of-flight (TOF) MR angiography or phase-contrast angiography sequences, they are nevertheless useful for an assessment of the cerebral arterial architecture. Since arterial information is extracted from the same data set, its geometric relationship to the venous vasculature is known precisely.
This additional vascular information may lead to a better understanding of pathologic conditions, such as arterial-venous diseases or lacunar infarcts (10). Figure 7 shows an example where SW imaging clearly demonstrates signal cancellation near the nidus of an AVM owing to deposits of paramagnetic blood products. Together with TOF MR angiography, the projected magnitude and SW images allow a detailed delineation of the arterial feeders and the venous drainage as well as the nidus of the AVM.
View larger version (119K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7a. AVM in a 57-year-old patient. TOF imaging (TR/TE = 37 msec/6.9 msec, 20° flip angle, voxel size of 0.41 x 0.41 x 0.9 mm) and SW imaging (TR/TE = 57 msec/40 msec, 20° flip angle, voxel size of 0.5 x 0.5 x 2 mm) were performed with a 1.5-T Magnetom Symphony (Siemens Medical Solutions). (a, b) MIP images obtained over 28 mm (a) and 10 mm (b) of TOF data show an AVM in the right parietal-occipital lobe. The AVM is fed by the middle meningeal artery and the distal network originating from the basilar and vertebral arteries. (c) MIP image of the magnitude data from SW imaging (same volume as in a) shows only the occipital feeding vessels as bright structures. However, their delineation is reduced compared with that in a and b owing to turbulent flow and increased susceptibility artifacts. (d) mIP image from SW data (volume coverage comparable with that in b) shows large amounts of blood products. (e) Zoomed mIP image obtained over 12 mm of median filtered SW data shows draining veins (arrows).
|
|
View larger version (125K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7b. AVM in a 57-year-old patient. TOF imaging (TR/TE = 37 msec/6.9 msec, 20° flip angle, voxel size of 0.41 x 0.41 x 0.9 mm) and SW imaging (TR/TE = 57 msec/40 msec, 20° flip angle, voxel size of 0.5 x 0.5 x 2 mm) were performed with a 1.5-T Magnetom Symphony (Siemens Medical Solutions). (a, b) MIP images obtained over 28 mm (a) and 10 mm (b) of TOF data show an AVM in the right parietal-occipital lobe. The AVM is fed by the middle meningeal artery and the distal network originating from the basilar and vertebral arteries. (c) MIP image of the magnitude data from SW imaging (same volume as in a) shows only the occipital feeding vessels as bright structures. However, their delineation is reduced compared with that in a and b owing to turbulent flow and increased susceptibility artifacts. (d) mIP image from SW data (volume coverage comparable with that in b) shows large amounts of blood products. (e) Zoomed mIP image obtained over 12 mm of median filtered SW data shows draining veins (arrows).
|
|
View larger version (132K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7c. AVM in a 57-year-old patient. TOF imaging (TR/TE = 37 msec/6.9 msec, 20° flip angle, voxel size of 0.41 x 0.41 x 0.9 mm) and SW imaging (TR/TE = 57 msec/40 msec, 20° flip angle, voxel size of 0.5 x 0.5 x 2 mm) were performed with a 1.5-T Magnetom Symphony (Siemens Medical Solutions). (a, b) MIP images obtained over 28 mm (a) and 10 mm (b) of TOF data show an AVM in the right parietal-occipital lobe. The AVM is fed by the middle meningeal artery and the distal network originating from the basilar and vertebral arteries. (c) MIP image of the magnitude data from SW imaging (same volume as in a) shows only the occipital feeding vessels as bright structures. However, their delineation is reduced compared with that in a and b owing to turbulent flow and increased susceptibility artifacts. (d) mIP image from SW data (volume coverage comparable with that in b) shows large amounts of blood products. (e) Zoomed mIP image obtained over 12 mm of median filtered SW data shows draining veins (arrows).
|
|
View larger version (151K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7d. AVM in a 57-year-old patient. TOF imaging (TR/TE = 37 msec/6.9 msec, 20° flip angle, voxel size of 0.41 x 0.41 x 0.9 mm) and SW imaging (TR/TE = 57 msec/40 msec, 20° flip angle, voxel size of 0.5 x 0.5 x 2 mm) were performed with a 1.5-T Magnetom Symphony (Siemens Medical Solutions). (a, b) MIP images obtained over 28 mm (a) and 10 mm (b) of TOF data show an AVM in the right parietal-occipital lobe. The AVM is fed by the middle meningeal artery and the distal network originating from the basilar and vertebral arteries. (c) MIP image of the magnitude data from SW imaging (same volume as in a) shows only the occipital feeding vessels as bright structures. However, their delineation is reduced compared with that in a and b owing to turbulent flow and increased susceptibility artifacts. (d) mIP image from SW data (volume coverage comparable with that in b) shows large amounts of blood products. (e) Zoomed mIP image obtained over 12 mm of median filtered SW data shows draining veins (arrows).
|
|
View larger version (131K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7e. AVM in a 57-year-old patient. TOF imaging (TR/TE = 37 msec/6.9 msec, 20° flip angle, voxel size of 0.41 x 0.41 x 0.9 mm) and SW imaging (TR/TE = 57 msec/40 msec, 20° flip angle, voxel size of 0.5 x 0.5 x 2 mm) were performed with a 1.5-T Magnetom Symphony (Siemens Medical Solutions). (a, b) MIP images obtained over 28 mm (a) and 10 mm (b) of TOF data show an AVM in the right parietal-occipital lobe. The AVM is fed by the middle meningeal artery and the distal network originating from the basilar and vertebral arteries. (c) MIP image of the magnitude data from SW imaging (same volume as in a) shows only the occipital feeding vessels as bright structures. However, their delineation is reduced compared with that in a and b owing to turbulent flow and increased susceptibility artifacts. (d) mIP image from SW data (volume coverage comparable with that in b) shows large amounts of blood products. (e) Zoomed mIP image obtained over 12 mm of median filtered SW data shows draining veins (arrows).
|
|
Phase Imaging
The phase represents the other half of the complete acquired signal information. Phase images are independent of the flip angle (35) and thus inhomogeneities (15). Since independent of B1 they provide additional anatomic information with a specific contrast, phase processing and phase image visualization are central aspects of GUIBOLD.
Anatomic phase imaging exploits susceptibility differences between tissues. Phase images allow differentiation between gray and white matter as well as between tissues with different iron content. Owing to the high sensitivity to local susceptibility changes, phase images can also demonstrate the iron distribution in tissue (eg, in the putamen) (36) and may thus improve diagnosis of neuroferritinopathies, such as Huntington disease, Alzheimer disease, or Parkinson disease (37). With GUIBOLD, mIPs and average projections (Fig 8) of unwrapped phase images can be caculated and displayed. These projections enhance borderlines of different tissues and veins. Even the deep structures of the basal ganglia (Fig 8b) and the corpus callosum (Fig 8a) are distinguishable in high detail.
View larger version (133K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8a. Anatomic phase images of a patient with multiple cavernomas. (a) Average projection image obtained over data from 20 mm of adjacent phase images shows good contrast of both gray and white matter. (b) Zoomed image of the region within the square in a shows the basal ganglia in more detail. The deep structure of the basal ganglia (putamen, thalamus, capsules, and caudate nucleus) is differentiable due to the different iron content. Heme iron inside venous vessels and cavernous lesions is also demonstrated. The two dark spots close to the basal ganglia are cavernous lesions. On average projection images obtained over the whole lesion, the signal is characterized by a dark spot surrounded by a bright rim (arrow).
|
|
View larger version (170K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8b. Anatomic phase images of a patient with multiple cavernomas. (a) Average projection image obtained over data from 20 mm of adjacent phase images shows good contrast of both gray and white matter. (b) Zoomed image of the region within the square in a shows the basal ganglia in more detail. The deep structure of the basal ganglia (putamen, thalamus, capsules, and caudate nucleus) is differentiable due to the different iron content. Heme iron inside venous vessels and cavernous lesions is also demonstrated. The two dark spots close to the basal ganglia are cavernous lesions. On average projection images obtained over the whole lesion, the signal is characterized by a dark spot surrounded by a bright rim (arrow).
|
|
Moreover, phase images can be used to differentiate between the magnetic type of nearly spherical lesions in brain tissue (whether they are of a paramagnetic or diamagnetic nature). This becomes possible by analyzing the signs of the resulting magnetic dipole pattern (Fig 9b). Figure 9a is a simulated phase image with a field inhomogeneity induced by a spherical paramagnet in a coronal view. The expected decreased phase at both lobes and the increased phase at the center of the perturbation are clearly visible. This typical paramagnetic pattern was also found in vivo resulting from a cavernous lesion (Fig 9c, 9d). A reversed field distribution, with higher phase at both lobes and lower phase at the equator, was observed for a diamagnetic calcified tumor (Fig 9e). Since SW imaging is usually performed in axial orientation, the characterization of the phase distribution resulting from a lesion on axial images is straightforward (39). GUIBOLD’s different visualization modes for magnitude and phase allow quick assessment of the field inhomogeneities produced by such lesions. Thus, the classification of nearly spherical lesions on the basis of phase images may help physicians in their final diagnosis and may also help avoid secondary computed tomography scans of patients.
View larger version (27K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9a. (a) Simulated coronal phase image shows a spherical paramagnetic field distribution (38). Phase is decreased at both lobes and increased at the equator. (b) Isosurface representation of the field distribution (with B0 parallel to the z axis) induced by a spherical perturber with a different susceptibility than its surroundings. The signs of the two lobes are reversed, compared with that of the equatorial rim. If the perturber is more paramagnetic, both lobes possess positive values and the equatorial rim is negative. This effect is reversed for diamagnetic perturbers. (c) Coronal view based on unwrapped phase images (acquired in axial orientation, reconstructed and visualized with GUIBOLD) shows a cavernous lesion with the typical paramagnetic pattern. (d) Isosurface rendering of the lesion in c. (e) Coronal phase image shows a calcified tumor. Calcified tissue reacts as a diamagnetic perturber and shows the typical diamagnetic dipole pattern, with the phase increased at the two lobes (arrowheads) and decreased at the equatorial rim (arrows).
|
|
View larger version (77K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9b. (a) Simulated coronal phase image shows a spherical paramagnetic field distribution (38). Phase is decreased at both lobes and increased at the equator. (b) Isosurface representation of the field distribution (with B0 parallel to the z axis) induced by a spherical perturber with a different susceptibility than its surroundings. The signs of the two lobes are reversed, compared with that of the equatorial rim. If the perturber is more paramagnetic, both lobes possess positive values and the equatorial rim is negative. This effect is reversed for diamagnetic perturbers. (c) Coronal view based on unwrapped phase images (acquired in axial orientation, reconstructed and visualized with GUIBOLD) shows a cavernous lesion with the typical paramagnetic pattern. (d) Isosurface rendering of the lesion in c. (e) Coronal phase image shows a calcified tumor. Calcified tissue reacts as a diamagnetic perturber and shows the typical diamagnetic dipole pattern, with the phase increased at the two lobes (arrowheads) and decreased at the equatorial rim (arrows).
|
|
View larger version (103K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9c. (a) Simulated coronal phase image shows a spherical paramagnetic field distribution (38). Phase is decreased at both lobes and increased at the equator. (b) Isosurface representation of the field distribution (with B0 parallel to the z axis) induced by a spherical perturber with a different susceptibility than its surroundings. The signs of the two lobes are reversed, compared with that of the equatorial rim. If the perturber is more paramagnetic, both lobes possess positive values and the equatorial rim is negative. This effect is reversed for diamagnetic perturbers. (c) Coronal view based on unwrapped phase images (acquired in axial orientation, reconstructed and visualized with GUIBOLD) shows a cavernous lesion with the typical paramagnetic pattern. (d) Isosurface rendering of the lesion in c. (e) Coronal phase image shows a calcified tumor. Calcified tissue reacts as a diamagnetic perturber and shows the typical diamagnetic dipole pattern, with the phase increased at the two lobes (arrowheads) and decreased at the equatorial rim (arrows).
|
|
View larger version (51K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9d. (a) Simulated coronal phase image shows a spherical paramagnetic field distribution (38). Phase is decreased at both lobes and increased at the equator. (b) Isosurface representation of the field distribution (with B0 parallel to the z axis) induced by a spherical perturber with a different susceptibility than its surroundings. The signs of the two lobes are reversed, compared with that of the equatorial rim. If the perturber is more paramagnetic, both lobes possess positive values and the equatorial rim is negative. This effect is reversed for diamagnetic perturbers. (c) Coronal view based on unwrapped phase images (acquired in axial orientation, reconstructed and visualized with GUIBOLD) shows a cavernous lesion with the typical paramagnetic pattern. (d) Isosurface rendering of the lesion in c. (e) Coronal phase image shows a calcified tumor. Calcified tissue reacts as a diamagnetic perturber and shows the typical diamagnetic dipole pattern, with the phase increased at the two lobes (arrowheads) and decreased at the equatorial rim (arrows).
|
|
View larger version (58K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9e. (a) Simulated coronal phase image shows a spherical paramagnetic field distribution (38). Phase is decreased at both lobes and increased at the equator. (b) Isosurface representation of the field distribution (with B0 parallel to the z axis) induced by a spherical perturber with a different susceptibility than its surroundings. The signs of the two lobes are reversed, compared with that of the equatorial rim. If the perturber is more paramagnetic, both lobes possess positive values and the equatorial rim is negative. This effect is reversed for diamagnetic perturbers. (c) Coronal view based on unwrapped phase images (acquired in axial orientation, reconstructed and visualized with GUIBOLD) shows a cavernous lesion with the typical paramagnetic pattern. (d) Isosurface rendering of the lesion in c. (e) Coronal phase image shows a calcified tumor. Calcified tissue reacts as a diamagnetic perturber and shows the typical diamagnetic dipole pattern, with the phase increased at the two lobes (arrowheads) and decreased at the equatorial rim (arrows).
|
|
However, for lesions with more complex shapes, the problem becomes more intricate and requires much more research on the induced phase changes and on the solution of the inverse problem to deduce the magnetic nature of a lesion for a given phase distribution. In addition, differentiation between two paramagnetic lesions (eg, hemorrhagic and melanotic lesions) also remains complicated, even if the proper magnetic susceptibility could be known, owing to pathologic variations within these lesions, such as the alterable melanin content of melanotic lesions (40).
|
Discussion
|
|---|
GUIBOLD fills the gap between researchers experienced in SW imaging data acquisition and reconstruction and clinical users, who so far have had to rely on the rather limited MR imaging system software. In SW imaging three images, magnitude, phase, and a combination of both, are produced. In particular, computation and assessment of phase images require high flexibility in data processing. This flexibility is warranted with GUIBOLD due to the possible adjustment of many parameters concerning k-space and SW imaging data processing. The computation time for a typical SW imaging data set with a matrix of 512 x 384 x 96 is approximately 60 seconds on a 2.8-GHz Linux workstation with 1 GB of random-access memory. This is fast enough for empirical adjustment and optimization of the reconstruction for a certain clinical application. On the other hand, a set of default parameters that represents the current state of the art has been provided (26,41) and allows the less experienced user to obtain good results for a wide range of applications.
As improved image processing methods for angiographic imaging are becoming more important, further segmentation and visualization methods may be integrated into GUIBOLD in the future. One feature could be the segmentation of venous vessels from SW imaging data (42,43) to create 3D models of the cerebral venous vasculature (44). Another important aspect is to integrate quantitative MR imaging methods and special statistical functions or texture analysis methods based on regions of interest (45). These features may be implemented via open-source toolkits available in various programming languages (eg, C, C++, or Java). For instance, further segmentation techniques of the C++ Insight Segmentation and Registration Toolkit (http://www.itk.org) can be made available for IDL via dynamic loadable modules (46) and then integrated into GUIBOLD.
One drawback of SW imaging is its inherent sensitivity to susceptibility artifacts close to air-bone (auditory channel) or air-tissue (paranasal sinuses) boundaries as well as to the presence of larger amounts of paramagnetic substances, such as hemosiderin or melanin. These susceptibility artifacts may result in severe signal cancellations in the magnitude images. Therefore, even with adjustment of the various phase unwrapping methods provided by GUIBOLD, incomplete phase correction may occur. Image interpretation and detection of lesions may then be hampered near such brain areas (47). Another disadvantage of SW imaging may be the rather long acquisition times of approximately 10 minutes at 1.5 T and therefore the sensitivity to motion artifacts, blurring, and ghosting. These patient motion artifacts should be avoided during data collection, since no motion artifact reduction is provided by GUIBOLD. Iterative autofocusing techniques have been presented to correct motion by means of k-space postprocessing based on magnitude entropy (48) or the squared normalized gradient magnitude of each pixel (49). However, the usefulness of these approaches for SW imaging still has to be evaluated.
In conclusion, GUIBOLD provides easy access to a wide range of research and clinical possibilities for SW imaging. It enables radiologists, medical technicians, and radiographers to explore the full potential and to quickly translate current research into improved diagnosis.
|
Conclusions
|
|---|
We have designed an imaging system–independent cross-platform stand-alone (off-line) application suite (GUIBOLD) for SW imaging data reconstruction with the ability to interactively set processing as well as display parameters. GUIBOLD is capable of presenting SW, magnitude, and unwrapped phase images in different specifiable orientations and projections (axial, sagittal, coronal) with various rendering techniques (MIP, mIP, and average projection). Region-of-interest analysis tools are also embedded in the application. The processing suite accepts different inputs (including the DICOM standard) and a variety of outputs (eg, DICOM, BMP, JPEG, TIFF, and PNG). The clinical potential of SW imaging has been demonstrated in several clinical applications, including visualization of multiple cavernomas, differentiation of hemorrhagic from calcified lesions by using phase images, and detailed delineation of AVMs. In addition, SW imaging helps to improve classification of tumors and assessment of traumas, cerebral hemorrhages, and venous thromboses. The SW imaging phase contrast may be used to visualize neuroanatomic details and to diagnose neuroferritinopathies. In summary, GUIBOLD enables fast and easy access to SW imaging for clinical routine and research purposes.
Note: GUIBOLD and a test data set are freely available over our Web page, http://www.mrt.uni-jena.de. Execution of GUIBOLD requires the freely available IDL virtual machine.
|
TAKE-HOME POINTS
|
|---|
- The IDL virtual machine makes it possible to run GUIBOLD as a cross-platform application (Windows [Microsoft, Redmond, Wash], Linux, and Mac OS X [Apple Computer, Cupertino, Calif]) without having to purchase an IDL license.
- Furthermore, system-independent DICOM images (in image space) with magnitude and phase information (wrapped or unwrapped after phase correction implemented on the particular MR system) are also supported.
- Evaluation and visualization of SW data with GUIBOLD allows one to differentiate between the arterial vessel architecture (as seen on the magnitude images) and the venous vessel architecture (as seen on SW images) (Fig 3).
- Phase images are independent of the flip angle (35) inhomogeneities (15). and thus independent of B1
- On the other hand, a set of default parameters that represents the current state of the art has been provided (26,41) and allows the less experienced user to obtain good results for a wide range of applications.
|
Footnotes
|
|---|
Abbreviations: AVM = arteriovenous malformation, BOLD = blood oxygen level–dependent, DICOM = Digital Imaging and Communications in Medicine, GUI = graphical user interface, MIP = maximum intensity projection, mIP = minimum intensity projection, SW = susceptibility-weighted, TE = echo time, 3D = three-dimensional, TOF = time of flight, TR = repetition time, 2D = two-dimensional
|
References
|
|---|
- Reichenbach JR, Haacke EM. High resolution BOLD venographic imaging: a window into brain function. NMR Biomed 2001;14(7–8):453–467.[CrossRef][Medline]
- Ogawa S, Lee T, Kay AR, Tank D. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci U S A 1990; 87(24):9868–9872.[Abstract/Free Full Text]
- Reichenbach JR, Essig M, Haacke EM, et al. High-resolution venography of the brain using magnetic resonance imaging. MAGMA 1998;6(1):62–69.[Medline]
- Tong KA, Ashwal S, Holshouser BA, et al. Hemorrhagic shearing lesions in children and adolescents with posttraumatic diffuse axonal injury: improved detection and initial results. Radiology 2003;227 (2):332–339.[Abstract/Free Full Text]
- Hermier M, Nighoghossian N. Contribution of susceptibility-weighted imaging to acute stroke assessment. Stroke 2004;35(8):1989–1994.[Abstract/Free Full Text]
- Haacke EM, Herigault G, Yu Y, et al. Observing tumor vascularity noninvasively using magnetic resonance imaging. Image Anal Stereol 2002;21: 107–113.
- Barth M, Nöbauer-Huhmann IM, Reichenbach JR, et al. High-resolution 3D contrast-enhanced BOLD MR-venography (CEMRV) of brain tumors at 3 Tesla: first clinical experience and comparison with 1.5 Tesla. Invest Radiol 2003;38(7):409–414.[CrossRef][Medline]
- Rauscher A, Sedlacik J, Fitzek C, et al. Investigation of tumor activity and anatomy by SWI during the inhalation of carbogen and air [abstr]. In: Proceedings of the 13th Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2005; 409.
- Sehgal V, Delproposto Z, Haddar D, et al. Susceptibility-weighted imaging to visualize blood products and improve tumor contrast in the study of brain masses. J Magn Reson Imaging 2006;24(1): 41–51.[CrossRef][Medline]
- Sehgal V, Delproposto Z, Haacke EM, et al. Clinical applications of neuroimaging with susceptibility-weighted imaging. J Magn Reson Imaging 2005; 22(4):439–450.[CrossRef][Medline]
- Rauscher A, Sedlacik J, Barth M, Haacke EM, Reichenbach JR. Noninvasive assessment of vascular architecture and function during modulated blood oxygenation using susceptibility weighted magnetic resonance imaging. Magn Reson Med 2005; 54(1):87–95.[CrossRef][Medline]
- Sedlacik J, Rauscher A, Deistung A, Kaiser WA, Reichenbach JR. Caffeine-induced contrast enhancement in volunteers by using susceptibility weighted imaging (SWI) [abstr]. Eur Radiol 2005; 15(suppl 1):341.
- Rauscher A, Barth M, Reichenbach JR, Stollberger R, Moser E. Automated unwrapping of MR phase images applied to BOLD MR-venography at 3 Tesla. J Magn Reson Imaging 2003;18(2):175–180.[CrossRef][Medline]
- Kholmovski E, Parker D. High resolution magnetic resonance venography at 3 Tesla: optimized acquisition, reconstruction, and post-processing [abstr]. In: Proceedings of the 14th Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2006; 810.
- Rauscher A, Sedlacik J, Barth M, Mentzel HJ, Reichenbach JR. Magnetic susceptibility-weighted MR phase imaging of the human brain. AJNR Am J Neuroradiol 2005;26(4):736–742.[Abstract/Free Full Text]
- Reichenbach JR, Barth M, Haacke EM, Klarhöfer M, Kaiser WA, Moser E. High-resolution MR venography at 3.0 Tesla. J Comput Assist Tomogr 2000;24(6):949–957.[CrossRef][Medline]
- Reichenbach JR, Venkatesan R, Schillinger D, Haacke EM. Small vessels in the human brain: MR venography with deoxyhemoglobin as an intrinsic contrast agent. Radiology 1997;204(1):272–277.[Abstract/Free Full Text]
- Noebauer-Huhmann IM, Pinker K, Barth M, et al. Contrast-enhanced, high-resolution, susceptibility-weighted magnetic resonance imaging of the brain: dose-dependent optimization at 3 tesla and 1.5 tesla in healthy volunteers. Invest Radiol 2006;41(3): 249–255.[CrossRef][Medline]
- Lin W, Mukherjee P, An H, et al. Improving high-resolution MR BOLD venographic imaging using a T1 reducing contrast agent. J Magn Reson Imaging 1999;10(2):118–123.[CrossRef][Medline]
- Research Systems. Getting started with IDL. Version 6.2. Boulder, Colo: Research Systems, 2005.
- Harris FJ. On the use of windows for harmonic analysis with the discrete Fourier transform. Proc IEEE 1978;66:51–83.
- Noll DC, Nishimura DG, Macovski A. Homodyne detection in magnetic resonance imaging. IEEE Trans Med Imaging 1991;10(2):154–163.[CrossRef][Medline]
- Witoszynskyj S, Rauscher A, Reichenbach JR, Barth M. UN (ase UNwrapping): validation of a 2D region-growing phase unwrapping program [abstr]. In: Proceedings of the 15th Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2007; 3436.
- Xu W, Cumming I. A region growing algorithm for InSAR phase unwrapping. IEEE Trans Geosci Remote Sensing 1999;37(1):124–134.[CrossRef]
- Witoszynskyj S, Rauscher A, Reichenbach JR, Barth M. Automated phase unwrapping of MR images at different field strengths using multiple quality maps [abstr]. In: Proceedings of the 13th Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2005; 439.
- Haacke EM, Xu Y, Cheng YC, Reichenbach JR. Susceptibility weighted imaging (SWI). Magn Reson Med 2004;52(3):612–618.[CrossRef][Medline]
- Research Systems. IDL reference guide. Version 6.0. Boulder, Colo: Research Systems, 2003; 2477–2482.
- Research Systems. IDL reference guide. Version 6.0. Boulder, Colo: Research Systems, 2003; 1810–1829.
- Analyze 7.5 file format. http://www.mayo.edu/bir/PDF/ANALYZE75.pdf.
- Essig M, Reichenbach JR, Schad LR, Schoenberg SO, Debus J, Kaiser WA. High-resolution MR venography of cerebral arteriovenous malformations. Magn Reson Imaging 1999;17(10):1417–1425.[CrossRef][Medline]
- Idbaih A, Boukobza M, Crassard I, Porcher R, Bousser MG, Chabriat H. MRI of clot in cerebral venous thrombosis: high diagnostic value of susceptibility-weighted images. Stroke 2006;37(4): 991–995.[Abstract/Free Full Text]
- Cakmak S, Hermier M, Montavont A, et al. T2*-weighted MRI in cortical venous thrombosis. Neurology 2004;63(9):1698.[Free Full Text]
- Leach JL, Strub WM, Gaskill-Shipley MF. Cerebral venous thrombus signal intensity and susceptibility effects on gradient recalled-echo MR imaging. AJNR Am J Neuroradiol 2007;28(5):940–945.[Abstract/Free Full Text]
- Alexander AL, Chapman BE, Tsuruda JS, Parker DL. A median filter for 3D fast spin echo black blood images of cerebral vessels. Magn Reson Med 2000;43(2):310–313.[CrossRef][Medline]
- Abduljalil AM, Schmalbrock P, Novak V, Chakeres DW. Enhanced gray and white matter contrast of phase susceptibility-weighted images in ultra-high-field magnetic resonance imaging. J Magn Reson Imaging 2003;18(3):284–290.[CrossRef][Medline]
- Krishnamurthy B, Haacke EM, Ayaz M, Harder S, Kido D, Kirsch W. Iron deposition in the putamen [abstr]. In: Proceedings of the 14th Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2006; 2656.
- Haacke EM, Cheng NY, House MJ, et al. Imaging iron stores in the brain using magnetic resonance imaging. Magn Reson Imaging 2005;23(1):1–25.[CrossRef][Medline]
- Schenck JF. The role of magnetic susceptibility in magnetic resonance imaging: MRI magnetic compatibility of the first and second kinds. Med Phys 1996;23(6):815–851.[CrossRef][Medline]
- Deistung A, Mentzel HJ, Rauscher A, Witoszynskyj S, Kaiser WA, Reichenbach JR. Demonstration of paramagnetic and diamagnetic cerebral lesions by using susceptibility weighted phase imaging (SWI). Z Med Phys 2006;16(4):261–267.[Medline]
- Isiklar I, Leeds NE, Fuller GN, Kumar AJ. Intracranial metastatic melanoma: correlation between MR imaging characteristics and melanin content. AJR Am J Roentgenol 1995;165(6):1503–1512.[Abstract/Free Full Text]
- Deistung A, Rauscher A, Sedlacik J, Witoszynskyj S, Reichenbach JR. Optimization of data processing in susceptibility-weighted imaging. Biomed Tech 2005;50(suppl 1 pt 2):1168–1169.
- Wang Y, Yu Y, Li D, et al. Artery and vein separation using susceptibility dependent phase in contrast-enhanced MRA. J Magn Reson Imaging 2000;12 (5):661–670.[CrossRef][Medline]
- Deistung A, Kocinski M, Materka A, Szczypinski P, Reichenbach JR. Segmentation of venous vessels using multi-scale vessel enhancement filtering in susceptibility weighted imaging [abstr]. In: Proceedings of the 14th Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2006; 1948.
- Kocinski M, Deistung A, Szczypinski PM, et al. Towards segmentation, visualization and quantification of vascular trees obtained from high-resolution susceptibility weighted MR imaging (SWI) and time-of-flight angiography of the human brain [abstr]. MAGMA 2006;19(suppl 7):46.[CrossRef][Medline]
- Hajek M, Dezortova M, Materka A, Lerski R. Texture analysis in magnetic resonance imaging. Prague, Czech Republic: Med4publishing, 2006.
- Research Systems. External development guide. Version 6.2. Boulder, Colo: Research Systems, 2005.
- Pinker K, Stavrou I, Szomolanyi P, et al. Improved preoperative evaluation of cerebral cavernomas by high-field, high-resolution susceptibility-weighted magnetic resonance imaging at 3 Tesla: comparison with standard (1.5 T) magnetic resonance imaging and correlation with histopathological findings—preliminary results. Invest Radiol 2007;42(6):346–351.[CrossRef][Medline]
- Atkinson D, Hill DL, Stoyle PN, et al. Automatic compensation of motion artifacts in MRI. Magn Reson Med 1999;41(1):163–170.[CrossRef][Medline]
- Lin W, Song HK. Improved optimization strategies for autofocusing motion compensation in MRI via the analysis of image metric maps. Magn Reson Imaging 2006;24(6):751–760.[CrossRef][Medline]
Top
Abstract
Introduction
