Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Magn Reson Imaging. Author manuscript; available in PMC 2010 November 18.
Published in final edited form as:
PMCID: PMC2987697

Simulation Study of Susceptibility Gradients Leading to Focal Myocardial Signal Loss



To assess the cause of a “bite”-shaped signal void artifact often seen in 1.5 Tesla (T) and 3T gradient echo MR images in myocardium along the infero-apical border of the heart, MRI simulation was used to conduct experiments impossible in reality. Two previous studies attempting to explain the origin of this artifact came to different conclusions. One suggested deoxygenated blood in the posterior vein of the left ventricle (PVLV) leads to a susceptibility gradient that causes the artifact. The other suggested the difference in susceptibility between lung tissue and myocardium was responsible. This study assessed the relative effect of each possible cause.

Materials and Methods

Anthropometric phantoms were developed for use with a previously reported MRI simulator. The images were simulated at 3T with gradient echo scans using TE = 4 ms, TR = 25 ms, and θ = 25°.


The simulations indicate that both susceptibility differences can lead to signal losses in the area of the artifact with contributions from the PVLV being more localized while lung tissue effects are stronger but more spatially distributed.


The data support the conclusion that both differences together, rather than one or the other, are responsible for the artifact.

Keywords: cardiac MRI, artifacts, distortions, magnetic field inhomogeneity, lung

MODERN CARDIAC MR imaging has shown the ability to reveal information beyond simple anatomy. Current techniques include evaluations of cardiac mass (14), measures of ejection fraction and contractility (5,6), as well as susceptibility-based contrast methods to assess myocardial perfusion and viability using exogenous contrast agents including gadolinium chelates (3,79) or by taking advantage of the BOLD effect using stress-inducing compounds such as dobutamine (10) or dipyridamole (1012). Several cardiac imaging groups have reported large static field inhomogeneities in and near the heart (1118) giving rise to a characteristic “bite”-shaped artifact. The artifact is seen as a highly focal, “bite”-shaped signal void along the infero-apical margin of the heart (Fig. 1). Pulse sequences that are highly sensitive to susceptibility differences such as many gradient echo methods, scans with echo planar readouts, and several other fast imaging techniques are most affected (14). The presence of this artifact can affect the interpretation of cardiac MR images.

Figure 1
Human cardiac MR image at 3T. TFE black blood technique. θ = 25°, TR = 25 ms, TE = 11.53 ms, FOV = 290 mm, BW = 543 Hz. This is the third image of a four image set used to map T2* decay. Note the loss of signal in the inferoapical region ...

Previous studies into the cause of this artifact have agreed that it is caused by magnetic susceptibility gradients but the relative influence of the various gradients around the heart has not been agreed upon. One study noted that the artifacts appeared near large epicardial veins, specifically the posterior vein of the left ventricle (PVLV) and, in some subjects, the great cardiac vein (17). The authors suggested that areas of high field inhomogeneity in measured field maps were generated by the difference in susceptibility between the deoxygenated blood flowing in large epicardial veins and the susceptibility of surrounding myocardium.

In another study, experiments with a porcine model were conducted to examine several possible causes for the artifact (14). The study showed that there was very little dependence of the artifact on lung volume. The authors did note there was a slight improvement in the end-inspiratory images and attributed this to a shift in the artifact to a more apical position. The authors found that the artifact was not dependent on the oxygenation of the blood in the coronary sinus (14), apparently contradicting the conclusions of the human study (17). The porcine study also showed that surgical resection of the lungs away from the myocardium removed the artifact. When the chest was opened, the lungs partially collapsed. Significant artifacts were still seen near the “points” of the now collapsed, triangular lungs. When the lungs were resected out of the chest, and the cavity was filled with copper sulfate solution, the artifacts were removed. Furthermore, the artifact was removed independent of the position of the diaphragm, reducing the likelihood that sub-diaphragmatic structures are responsible. The authors of this study concluded lung tissue is the most likely source of the susceptibility artifacts although they noted that structural differences in heart and lung anatomy between pigs and humans prevent making this conclusion unequivocally for humans (14). An earlier report by the same group showed that free air in the chest after lung resection also did not produce the artifact, leading to the conclusion that the packaging of air into alveoli and small bronchioles rather than the extreme susceptibility difference between air and soft tissue is the source of the artifact (13).

The results and conclusions of the previous studies (13,14,17) lead to some interesting questions. If the large epicardial veins are the primary cause of the artifact, as suggested in the human study, why did increasing the oxygen saturation of the vessels and thus reducing the susceptibility difference between the blood and the myocardium not lead to a reduction in the artifact in the porcine study? Second, if the heart–lung interface is the primary source of the susceptibility gradient, why does the artifact appear so focally rather than along the entirety of the heart–lung interface?

It is the purpose of this study to assess by means of rigorous simulations the specific contributions of the PVLV and coronary sinus (CS) versus the heart–lung interface to the formation of static field inhomogeneities in the heart and thus imaging artifacts, with particular emphasis on the focal “bite”-shaped artifact. Simulating experiments allows control over the experiment such as setting blood oxygenation levels, changing the susceptibility of an entire organ, and eliminating motion, things not possible in real life. The simulations also provide the ability to completely remove or change internal structures without disrupting the rest of the body as would be required with surgical preparation of an in vivo study.


Data generation was conducted in four discrete steps. First, the object to be imaged was defined in terms of tissue types and morphology. Second, the digital object definition was converted to a matrix of susceptibility values. Third, the susceptibility matrix was used to calculate static magnetic field perturbations. Maps of the static field perturbations were calculated over a range of values for the susceptibility of blood in the PVLV (χPVLV). For each value of χPVLV, the field maps were calculated twice, once with normal left lung susceptibility, and once with the susceptibility of the left lung altered to equal that of soft tissue. Finally, four of these calculated field maps were selected and used in combination with the original object definition to create representative images using an image simulator. All preparations of the object definition were carried out in Matlab 7.2 (The MathWorks, Natick, MA) with the Image Processing Toolbox 5.2 (The MathWorks).

A digital object definition of the human male torso was created for use in MRI simulations by modifying the phantom created by Zubal et al. (19) Zubal and colleagues created the original phantom from 129 transverse computed tomography (CT) image slices that were manually segmented into the major internal structures of the body. The chest section of the Zubal phantom is a 192 × 96 × 78 voxel volume with isotropic 4-mm voxels and spans from the neck to slightly inferior of the diaphragm. Each voxel contained an integer representing a specific tissue type. The original Zubal phantom did not differentiate between oxygenated and deoxygenated blood. The object definition for the present report was created by modifying the Zubal phantom to contain oxygenated blood in the aorta, other arteries, and the left side of the heart and deoxygenated blood in the systemic veins and right side of the heart. Additionally, the large voxels were interpolated to 384 × 192 × 78, to represent a more realistic size for MRI. The new voxel dimensions were 2mm in-plane and remained 4 mm in the through-plane direction. Nearest-neighbor interpolation was used to maintain the integer data-type of the phantom. The phantom was also smoothed at this phase by taking the mode of values of a 3 × 3 kernel. This was done to remove the “blockiness” of the phantom following the interpolation. Additionally, a PVLV and CS were added to the Zubal phantom and initially treated as containing deoxygenated blood to complete the object definition. The diameter of the inserted PVLV and CS was a single voxel (2 mm), corresponding with the assumption of a 1 mm radius used in the previous human study (17).

The object definition of tissue types was converted to a 3D matrix of susceptibility values, referred to as susceptibility voxels, using a lookup table for all tissues other than the left lung and PVLV. The susceptibilities used for each tissue type are shown in Table 1. The susceptibility of bulk lung tissue was calculated as a weighted average of the susceptibilities of soft tissue and air. Because air has a near-zero susceptibility, multiplying the susceptibility of general soft tissue (–9.05 ppm) by 0.26, the density of inflated lung tissue (20), provides a reasonable estimate of the susceptibility for bulk lung tissue. The susceptibility of the left lung was set to either that of normal lung tissue or normal nonlung soft tissue, to simulate removing that lung's effects. The susceptibility of blood in the PVLV was varied from –9.2*10–6 to –7.6*10–6. This range includes the susceptibilities of both oxygenated and fully deoxygenated blood as calculated using equations from Spees et al. (21) assuming a hematocrit of 40%. The susceptibility was varied over this range to assess the effect of oxygenation of blood in the PVLV on susceptibility gradients in that region.

Table 1
Tissue Parameters Used for Simulation*

Maps of the static field perturbations were calculated using the Susceptibility-Voxel Convolution (SVC) method described by Yoder et al. (22) The SVC method takes advantage of the fact that the boundary element method can be rewritten as the convolution of the susceptibility voxels and a special kernel. While calculating the convolution directly is computationally expensive, Yoder exploits the fast Fourier transform and performs the convolution as a multiplication in reciprocal (k) space, thus improving the computational efficiency of the field map calculation.

Next, to focus on the area in question and reduce computational intensity, a targeted volume of interest (VOI) was selected from the object definition and corresponding field maps. This volume included five slices, each 4 mm thick, of the left ventricle. Each axial slice from the 128 × 128 × 5 voxel volume of interest contained the entire heart area within that slice, left lung, most of the right lung, and some surrounding tissues. The VOI was selected to include portions of the PVLV and CS. It is important that this selection of a VOI for the imaging simulator occur after the field map is calculated so that other thoracic and abdominal structures not contained in the VOI still affect the field map just as they would in a real situation. The VOI can then be used for imaging while still accounting for other structures outside the VOI. This is somewhat analo gous to restricting the field of view during actual imaging with the exception that aliasing will not occur with this simulation approach. Note that because the calculation of field offsets was completed before selection of the VOI, field effects from structures outside this volume are accounted for even though those structures are not included in the image simulation.

The MRI signal was generated using the simulator developed by Yoder et al. (22) The simulator is based on the Bloch equations and allows for inclusion of the previously calculated field offsets and consequent intravoxel dephasing and signal mismapping throughout acquisition protocols. The simulator takes as input the object definition, tissue-type information (T1, T2, spin density from Table 1), the calculated static field perturbations, and a user-defined pulse sequence. The signal generated is output as the resultant magnitude image. In all simulations, a standard, 2D, spoiled gradient echo pulse sequence was used (flip angle = 25 deg, TR = 25 ms, TE = 4 ms, BW = 543 Hz, FOV = 256 mm) with static field strength of 3T. A powerful feature of this simulator is that the dephashing effects exerted on one voxel are affected by adjacent voxels in both in- and through-plane directions and are calculated from first principles. This, when combined with the user-provided T2 information, accurately simulates T2* effects.

Four simulations were run to examine the effects of both the PVLV and left lung on image signal in the region of the artifact. To test the hypothesis that deoxygenated blood in the PVLV and CS is the primary cause of the artifact, images were simulated using field maps calculated with these structures containing oxygenated and deoxygenated blood. To test the hypothesis that the heart–lung interface is primarily responsible for the artifact, images were simulated using field maps calculated with the left lung tissue as normal and also with its susceptibility changed to the susceptibility of myocardium, thus removing the susceptibility gradient at the heart–“lung” interface. This models the experiment by Atalay et al. (14) in which the left lung was removed and replaced with a CuSO4 solution except that in this experiment T1 and T2 for the lungs will not change. By simply changing the susceptibility rather than replacing the tissue, any changes in image contrast can be attributed to the change in susceptibility and are not confounded by changes in T1 and T2. No changes were made to the right lung tissue in any simulation.

Analysis of the data was performed both at the level of the calculated field maps and at the level of the simulated images. For the calculated field maps, three regions of interest (ROIs) were selected: one around the PVLV, one along the heart-lung interface, and one in the background (Fig. 2). The mean field offset ± one standard deviation was calculated for each ROI and plotted against the values of χPVLV (Fig. 3). Additionally, the change in field offset compared with χPVLV was analyzed at each individual voxel. A linear regression was used to calculate a line predicting field offset vs. χPVLV for each individual voxel. The slope of this line was plotted against the position of the voxel. This created a new map that highlights areas most affected by changes in χPVLV (Fig. 4). For the simulated images (Fig. 5), the same three ROIs were selected (Fig. 2). Comparisons were made both by visual inspection and by calculating mean signal values in each ROI (Fig. 6).

Figure 2
ROIs used for analysis of simulated field maps and images.
Figure 3
Plots of mean field offset ±1 SD versus χPVLV for the three ROIs in Figure 2. Red lines are for simulations with normal lung susceptibility. Blue lines are for simulations with lung susceptibility changed to that of soft tissue. A: ROI ...
Figure 4
Map of the dependence of field variation on χPVLV as quantified by the slope of the field variation versus χPVLV at each point overlaid with an outline of the original phantom. Gray areas indicate no dependence of field variation on χ ...
Figure 5
Simulated Images. A–D: Deoxygenated blood in the PVLV and CS and normal lung susceptibility (A), deoxygenated blood in the PVLV and CS and altered lung susceptibility (B), oxygenated blood in the PVLV and CS and normal lung susceptibility (C), ...
Figure 6
Mean signal values from simulated images for the ROIs shown in Figure 2. Note the slightly lower and more varied signal around the PVLV when it is filled with deoxygenated blood.


The plots of mean field offset in each ROI are shown in Figure 3. The variation in χPVLV has no effect in the background and heart–lung interface regions, and only a slight effect in the region surrounding the PVLV. Changes in the susceptibility of the left lung had much larger effects, and affected all three regions. Mapping the magnitude of the effects led to clear areas where the variation in χPVLV caused the most change in the field offset, as shown in Figure 4.

The representative images from each simulation (the same slice level as the field maps used to create Fig. 4) are shown in Figure 5. Regions of local signal loss are seen in images with deoxygenated blood in the PVLV and CS (arrowhead, Fig. 5A,B) and the affected area is reduced in the simulated images with oxygenated blood in the PVLV and CS (arrowhead, Fig. 5C,D). Similarly, distributed signal loss is seen along much of the heart– lung interface in images using the normal susceptibility value of lung tissue (arrows, Fig, 5A,C). When the susceptibility of the left lung was changed to equal that of soft tissue, the distributed signal loss disappears (arrows, Fig. 5B,D).

Mean signal values for the three ROIs in each simulation are shown in Figure 6. For the ROI near the PVLV, signal changes are dependent on both the oxygenation (and, thus, susceptibility) of blood in the PVLV and CS and the susceptibility of lung tissue. The mean signal value for the ROI near the PVLV is 3.5% lower when the blood in the PVLV and CS is deoxygenated compared with oxygenated in images with normal lung susceptibility. This figure drops to 2.2% for images with altered left lung susceptibility. The standard deviation of the signal in this ROI is increased when the blood in the PVLV is deoxygenated rather than oxygenated by 8.8% and 10.3% for normal and altered left lung susceptibility, respectively. The mean signal value for the same ROI is 6.7% lower when the left lung susceptibility is normal than when the susceptibility is altered for images with deoxygenated blood in the PVLV and CS, and the standard deviation is increased by 16%. When the blood in the PVLV and CS is oxygenated, however, the change in mean signal between images with altered and normal left lung susceptibility falls to 5.5%, and the difference in standard deviation in signal values becomes 17%. For the ROI near the heart–lung interface, the mean signal decreases by 0.91% when the susceptibility of lung tissue is normal compared altered to be that of soft tissue, regardless of the oxygenation of blood in the PVLV. For the same manipulation, the standard deviation increased by 475% for deoxygenated blood in the PVLV and 450% for oxygenated blood in the PVLV. Altering the oxygenation of blood in the PVLV and CS did not lead to changes in the mean signal for the ROI along the heart–lung interface. The oxygenation of blood affected the standard deviation of signal in this ROI by 4.5%, but this effect was only present when the susceptibility of lung tissue was unaltered.


The authors of the previous human study (17) showed, by experiment, that the area surrounding the epicardial veins experiences increased T2* decay and signal loss. By means of theoretical arguments, they showed that the susceptibility difference between deoxygenated blood and myocardium could lead to increased signal loss. The authors of the porcine study concluded the draining veins are not primarily responsible after testing the effects of blood oxygenation in only two animals. It is possible this is not a representative population. Additionally, blood oxygenation was measured using a catheter in the coronary sinus, which contains mixed blood from a large number of veins. It is possible the oxygenation of the blood in the coronary sinus is not the same as the oxygenation of blood in the PVLV. The other evidence cited by the authors of the porcine study to conclude that the PVLV was not primarily responsible was the movement of the artifact at different stages of respiration despite what was claimed to be unchanged myocardial position. However, it is known that the heart moves with respiration in humans (23), and it is reasonable to assume the same occurs in pigs. It is entirely possible that respiratory motion of the heart resulted in slightly different imaging planes in which the myocardial shape appears very similar. Thus, motion of the heart (and its PVLV) due to respiration could explain the apparent motion of the artifact if the PVLV is indeed responsible.

In vivo, the heart is located between the lungs, diaphragm, and anterior chest wall. The interface between the heart and the lungs is extensive, yet the artifact in question appears at a specific location along the interface. Furthermore, if the air in the lungs is the source of the susceptibility gradient, replacing the heart–lung interface with free air should not eliminate the artifact, contrary to the experimental results (13). The authors claim the packaging of air into alveoli and small airways rather than the bulk effect of air in the thorax is the cause of the artifact. It seems unlikely that packaging of the air into small alveoli would limit signal loss to such a focal area, effectively preventing it elsewhere. It seems more likely that perturbations from air packaged into alveoli would lead to a distributed artifact along the heart–lung interface.

In this study, calculated field maps and image simulations were used to show the independent and combined effects of blood oxygenation in the PVLV and CS and the susceptibility of left lung tissue on the presence of susceptibility artifacts in the heart, especially near the PVLV. The susceptibility difference between deoxygenated blood and myocardium led to a local signal loss of 3.5% around the PVLV compared with the image with oxygenated blood in the PVLV and CS for the parameters tested. While this by itself is unimpressive, when taken with the change in variation of the signal values, this represents a noticeable change in image appearance as can be seen by visual inspection of the images. The increase in variation between the pixels in this ROI, as indicated by the change in standard deviation, is caused by an increase in the geometric distortions around the PVLV. The Jacobian term of the geometric distortion of the image described by Chang and Fitzpatrick (24) determines how the susceptibility gradients alter the apparent spatial localization of the signal without altering the overall strength of the signal. Signal that is incorrectly localized, but still localized within the chosen ROI, will not affect the mean signal value for the ROI, but may affect the variation of signal values and, thus, image appearance for that ROI. Independently, susceptibility gradients also lead to dephasing in gradient echo scans, in which case, only signal losses, never increases, are seen. While the susceptibility difference between the left lung tissue and myocardium also affected the signal from the region surrounding the PVLV, it was also found to cause a band of signal loss along the heart–lung interface. Finally, the data for the ROI surrounding the PVLV show the relative effects of both the oxygenation of blood in the PVLV and susceptibility of lung tissue on mean signal are dependent on the status of the other variable. The effects of blood oxygenation are greater when lung tissue has its normal susceptibility. Similarly, the effect of lung susceptibility is greatest when the blood in the PVLV and CS is completely deoxygenated. These results indicate that the focal artifact near the PVLV is likely caused by a combination of effects from susceptibility differences between deoxygenated blood in the PVLV and CS and the myocardium and susceptibility differences between the heart and left lung.

This study showed that the oxygenation of blood in the PVLV does affect signal from the myocardium in its immediate vicinity in the heart. Simulated images do show an artifact near the PVLV (arrowhead, Fig. 5) similar to those described in previous studies (13,14,17), and the change in signal accompanying a change in blood oxygenation is in agreement with the theory laid out by Kennan et al. (25) and cited in the human study (17). The present study also showed the large susceptibility difference between the lung tissue and myocardium leads to a significant artifact (arrows, Fig. 5). As expected, the artifact spans a large portion of the heart–lung interface, including the area near the PVLV. This is in contrast to the porcine study's conclu sion that the heart–lung interface is responsible for the focal artifact in that the effect of the heart–lung interface appears to be more widespread. However, it should be noted that the simulation assumes that the lungs are homogeneous structures, yet real lungs are divided into discrete sections with differing lobar structures for humans and pigs (14,26). The fissures between sections of lung may possibly contribute to a focusing of the signal loss due to variation in the geometry and tissue involvement of the heart–lung interface at specific locations along the myocardium of humans and pigs.

It is possible the myocardial signal loss due to the heart–lung interface often goes unnoticed due to its proximity to the already low signal from lung tissue adjacent to the heart. This peripheral myocardial signal loss may possibly lead to an underestimation of myocardial mass (27), especially along the lung margins.

The use of the simulator developed by Yoder et al. (22) has some limitations. The simulator is only designed to handle up to eight tissue types. Thus, for our object definition some structures in the upper abdomen had to be grouped together into a general soft tissue category. However, because the susceptibilities of soft tissues are very similar, this does not affect the presence of artifacts. At worst, this generalization leads to unrealistic image contrast for some tissues. In limiting the volume for which actual MR signal was simulated, the diaphragm was not included within the imaging volume. While susceptibility effects from the diaphragm and superior aspects of subdiaphragmatic surfaces were accounted for, the effect of the unique anatomic geometry at the interface of the heart, lung, and diaphragm was not directly investigated. Additionally, the original Zubal phantom as well as our object definitions did not account for partial volumes at tissue interfaces. Furthermore, internal variations in magnetic properties of any organ with the exception of the heart were not considered. Thus, the microstructure of the lungs was not provided to the simulator, and instead they were considered homogenous organs. Because T2* effects from the lung microstructure were not simulated, a T2* value was provided for image generation rather than T2. Finally, the simulator does not allow for oblique imaging planes, so the short-axis images in which these artifacts are normally seen could not be recreated. However, the axial plane analyzed was chosen because the axis of the PVLV is close to perpendicular to the imaging plane, as it would be on a short axis image of the heart.

The use of the simulator also provides several advantages. Using a digital object definition allows for modifications that would otherwise be impossible, such as the change in lung tissue susceptibility without changing T1 and T2 and without opening the chest. Use of the digital object definition also assures identical positioning for each experiment, removing the need for image registration or other compensation for between-study tissue displacements. Additionally, the simulator allows for the use of a “perfect” imaging system: images are simulated with a homogeneous applied field with field perturbations computed by the user, pulse sequences have homogeneous RF pulses and perfect spoiling, gradients used in the simulation are perfectly linear and have zero rise time, the effects of eddy currents and dynamic properties, such as blood flow, breathing, and heartbeats, are eliminated, and the system is free of noise. Thus, artifacts due solely to static inhomogeneities in the magnetic field are revealed.

In using a simulator that assumes perfectly homogenous RF pulses, the possibility that RF inhomogeneity may also contribute significantly to this artifact is ignored. Indeed, significant RF current density gradients have been shown to cause signal voids within the heart at high field strengths (28). Despite this possibility, our results indicate that susceptibility effects can explain this artifact.

Due to the highly focal nature of this artifact and the large susceptibility difference between lung tissue and myocardium, shimming methods have had little utility in attempting to reduce or eliminate the artifact (14). However, several methods of reducing this artifact have been suggested including using thinner slices reoriented to be orthogonal to the greatest susceptibility gradient, thus reducing the in-plane mismapping of signal due to susceptibility but also reducing the signal-to-noise ratio due to through-plane signal losses. Shortening TE, as is common for anatomic and functional images, reduces signal loss due to susceptibility gradients. However, TE cannot be shortened when T2*-dependent tissue contrast is needed or when quantifying T2*. Empirically, positioning the subject prone was reported to accomplish some artifact reduction (11).

In conclusion, for the parameters tested, we have shown that susceptibility differences caused by deoxygenated blood contained in the PVLV and CS and the heart–lung interface lead to artifacts presenting as myocardial signal voids in simulated images of the human chest. The data support the conclusion that the susceptibility of deoxygenated blood in the PVLV and CS is responsible for the focal nature of the “bite”-shaped myocardial signal void, with the susceptibility gradient due to the heart–lung interface being a large, but more distributed, contributor to the artifact.


The authors thank Craig Lorang for his early work on this research. The authors also thank Dr. Ning Xu and Dr. J. Michael Fitzpatrick for their collaboration and assistance working with the MRI simulator.


1. Alfakih K, Reid S, Jones T, Sivananthan M. Assessment of ventricular function and mass by cardiac magnetic resonance imaging. Eur Radiol. 2004;14:1813–1822. [PubMed]
2. Fogel MA. Assessment of cardiac function by magnetic resonance imaging. Pediatr Cardiol. 2000;21:59–69. [PubMed]
3. Heatlie GJ, Pointon K. Cardiac magnetic resonance imaging. Post-grad Med J. 2004;80:19–22. [PMC free article] [PubMed]
4. Pons-Llado G. Assessment of cardiac function by CMR. Eur Radiol. 2005;15(Suppl 2):B23–B32. [PubMed]
5. Constantine G, Shan K, Flamm SD, Sivananthan M. Role of MRI in clinical cardiology. Lancet. 2004;363:2162–2171. [PubMed]
6. Setser RM, Fischer SE, Lorenz CH. Quantification of left ventricular function with magnetic resonance images acquired in real time. J Magn Reson Imaging. 2000;12:430–438. [PubMed]
7. Ablitt NA, Gatehouse PD, Firmin DN, Yang GZ. Respiratory reordered UNFOLD perfusion imaging. J Magn Reson Imaging. 2004;20:817–825. [PubMed]
8. Earls JP, Ho VB, Foo TK, Castillo E, Flamm SD. Cardiac MRI: recent progress and continued challenges. J Magn Reson Imaging. 2002;16:111–127. [PubMed]
9. Haacke EM, Li D, Kaushikkar S. Cardiac MR imaging: principles and techniques. Top Magn Reson Imaging. 1995;7:200–217. [PubMed]
10. Pujadas S, Reddy GP, Weber O, Lee JJ, Higgins CB. MR imaging assessment of cardiac function. J Magn Reson Imaging. 2004;19:789–799. [PubMed]
11. Niemi P, Poncelet BP, Kwong KK, et al. Myocardial intensity changes associated with flow stimulation in blood oxygenation sensitive magnetic resonance imaging. Magn Reson Med. 1996;36:78–82. [PubMed]
12. Wacker CM, Bock M, Hartlep AW, et al. Changes in myocardial oxygenation and perfusion under pharmacological stress with dipyridamole: assessment using T2* and T1 measurements. Magn Reson Med. 1999;41:686–695. [PubMed]
13. Atalay MK, Poncelet BP, Kantor HL, Brady TJ, Weisskoff RM. Susceptibility artifact at the cardiac apex of porcine hearts at 3T is primarily due to the heart-lung interface.. Proc. ISMRM, 6th Annual Meeting; Sydney. 1998. p. 130.
14. Atalay MK, Poncelet BP, Kantor HL, Brady TJ, Weisskoff RM. Cardiac susceptibility artifacts arising from the heart-lung interface. Magn Reson Med. 2001;45:341–3435. [PubMed]
15. Jaffer FA, Wen H, Balaban RS, Wolff SD. A method to improve the B0 homogeneity of the heart in vivo. Magn Reson Med. 1996;36:375–383. [PubMed]
16. Nayak KS, Cunningham CH, Santos JM, Pauly JM. Real-Time Cardiac MRI at 3 Tesla. Magn Reson Med. 2004;51:655–660. [PubMed]
17. Reeder SB, Faranesh AZ, Boxerman JL, McVeigh ER. In vivo measurement of T*2 and field inhomogeneity maps in the human heart at 1.5 T. Magn Reson Med. 1998;39:988–998. [PMC free article] [PubMed]
18. Schar M, Kozerke S, Fischer SE, Boesiger P. Cardiac SSFP imaging at 3 Tesla. Magn Reson Med. 2004;51:799–806. [PubMed]
19. Zubal G, Harrell CR, Smith EO, Smith AL. Two dedicated software, voxel-based, anthropomorphic (torso and head) phantoms. [November 2006].
20. ICRU Tissue substitutes in radiation dosimetry and measurement. 1989.
21. Spees WM, Yablonskiy DA, Oswood MC, Ackerman JJ. Water proton MR properties of human blood at 1.5 Tesla: magnetic susceptibility, T1, T2, T2*, and non-Lorentzian signal behavior. Magn Reson Med. 2001;45:533–542. [PubMed]
22. Yoder DA, Zhao Y, Paschal CB, Fitzpatrick JM. MRI Simulator with object-specific field map calculations. Magn Reson Imaging. 2004;22:315–328. [PubMed]
23. Wang Y, Riederer SJ, Ehman RL. Respiratory motion of the heart: kinematics and the implications for the spatial resolution in coror-nary imaging. Magn Reson Med. 1995;33:713–719. [PubMed]
24. Chang H, Fitzpatrick JM. A Technique for accurate magnetic resonance imaging in the presence of field inhomogeneities. IEEE Trans Med Imaging. 1992;11:319–329. [PubMed]
25. Kennan RP, Zhong J, Gore JC. Intravascular susceptibility contrast mechanisms in tissues. Magn Reson Med. 1994;31:9–21. [PubMed]
26. Tirabassi M, Banever G, Tashjian D, Moriarty K. Quantification of lung sealing in the survival swine model. J Pediatr Surg. 2004;39:387–390. [PubMed]
27. Spuentrup E, Schroeder J, Mahnken AH, et al. Quantitative assessment of left ventricular function with interactive real-time spiral and radial MR imaging. Radiology. 2003;227:870–876. [PubMed]
28. Vaughan JT, Adriany G, Snyder CJ, et al. Efficient high-frequency body coil for high-field MRI. Magn Reson Med. 2004;52:851–859. [PubMed]
29. Schenck JF. The role of magnetic susceptibility in magnetic resonance imaging: MRI magnetic compatibility of the first and second kinds. Med Phys. 1996;23:815–850. [PubMed]
30. Sumanaweera TS, Glover GH, Binford TO, Adler JR. MR susceptibility misregistration correction. IEEE Trans Med Imaging. 1993;12:251–259. [PubMed]
31. Silvennoinen M, Clingman C, Golay X, Kauppinen R, van Zijl P. Comparison of the Dependence of Blood R2 and R2* on Oxygen Saturation at 1.5 and 4.7 Tesla. Magn Reson Med. 2003;49:47–60. [PubMed]
32. Hatabu H, Alsop DC, Listerud J, Bonnet M, Gefter WB. T2* and proton density measurement of normal human lung parenchyma using submillisecond echo time gradient echo magnetic resonance imaging. Eur J Radiol. 1999;29:245–252. [PubMed]
33. Collins DL, Zijdenbos AP, Kollokian V, et al. Design and construction of a realistic digital brain phantom. IEEE Trans Med Imaging. 1998;17:463–468. [PubMed]