Imaging and Sizing of Atrial Septal Defects by Magnetic Resonance
Background Development of techniques for percutaneous closure of atrial septal defects (ASDs) makes accurate noninvasive sizing of ASDs important for appropriate patient selection.
Methods and Results Magnetic resonance (MR) images of ASDs were obtained in 30 patients (mean age, 41±16 years) by both spin-echo and phase-contrast cine MR imaging. Spin-echo images were obtained in two orthogonal views (short-axis and four-chamber) perpendicular to the plane of the ASD. Spin-echo major and minor diameters were measured, and spin-echo defect area was calculated. Phase-contrast cine MR images were obtained in the plane of the ASD, and cine major diameter and defect area were measured from the region of signal enhancement or phase change due to shunt flow across the defect. MR measurements were compared with templates cut during surgery to match the defect or with ASD diameter determined by balloon sizing at catheterization. ASD size measured from cine MR images (y) agreed closely with catheterization and template standards (x). For major diameter, y=0.78x+5.7, r=.93, and SEE=3.4 mm. On average, spin-echo measurements overestimated major diameter and area of secundum ASDs by 48% and 125%, respectively.
Conclusions Phase-contrast cine MR images acquired in the plane of an ASD define the defect shape by the cross section of the shunt flow stream and allow noninvasive determination of defect size with sufficient accuracy to permit stratification of patients to closure of the defect by catheter-based techniques versus surgery. Spin-echo images, on the other hand, are not adequate for defining ASD size, because septal thinning adjacent to a secundum ASD may appear to be part of the defect.
As improved methods for nonsurgical closure of ASDs become available, an accurate noninvasive technique for sizing ASDs will be of value for selecting patients for this procedure. Such a technique will be particularly helpful when a patient with an uncomplicated ASD is a candidate to go directly to surgery without catheterization if the defect size is too large to permit percutaneous closure. Appropriate selection of patients for percutaneous closure also requires evaluation of the proximity of the defect to other cardiac structures such as the atrioventricular valves, the venae cavae, and the pulmonary veins and exclusion of the presence of multiple defects or the presence of associated lesions such as anomalous pulmonary venous drainage.
Visualization of ASDs by MRI with the spin-echo technique is well established,1 as is the use of dynamic imaging with gradient-echo cine techniques to visualize ASD shunt flow.2 Investigators have also compared the apparent ASD size in spin-echo images with measurements obtained during surgery.3 4
Since our early experience, we have seen examples in which spin-echo measurements of ASD diameter gave misleading results. Here we describe and evaluate a different phase-contrast cine MRI method for determining ASD location, shape, and dimensions. A comparative study of these two MRI techniques, with reference to size data obtained during surgery or in the catheterization laboratory, has not been reported.
Fifty-one consecutive patients with a known or suspected ASD underwent an MRI study to determine the defect size and location. In one patient, the septum was intact by both MRI and ultrasound (including transesophageal echo). The other 50 patients had either an ASD or a PFO. Two patients had incomplete MRI data for sizing. In 18 patients, no correlative direct measurements of the defect could be obtained because catheterization or surgery was not clinically indicated, the patient chose not to proceed, the surgery was done elsewhere, or planned catheterization for percutaneous ASD closure awaited availability of a modified device. The MRI sizing technique was validated against an independent standard in the remaining 30 patients, who are the subject of this report. Their average age was 41±16 years (range, 10 to 67 years). The average time between MRI and catheterization or surgery was 42±79 days (range, 0 to 270 days; <2 weeks in 20 patients). All patients were informed about the purpose of the study and agreed to the study, which was performed in accordance with protocol guidelines reviewed and approved by the subcommittee on human studies at our hospital.
Magnetic Resonance Imaging
The MRI studies were performed on a Technicare system operating at 0.6 T. The cine images were acquired with a pulse sequence that had been modified in house from the Technicare source code to allow velocity encoding or compensation. The last two patients were studied with a 1.5-T GE Signa scanner.
Multislice spin-echo T1-weighted diagnostic images were obtained in the four-chamber view (Fig 1A⇓) as well as in a modified short-axis view (defined in the legend of Fig 1⇓). The slice positions, acquisition order, and gate delay from the R wave for this short-axis view were adjusted to ensure that the slice through the middle of the ASD (Fig 1B⇓) was acquired during the early diastolic filling period when ASD shunt flow was maximum (≈400 ms after the R wave). This allowed optimal planning from this short-axis slice of subsequent cine imaging of the shunt flow stream at the orifice by accurately localizing the ASD plane in space at the time of peak flow.
Three immediately adjacent (no gap) or slightly (25%) overlapping cine MRI acquisitions were obtained sequentially, with the middle slice prescribed in the plane of the ASD. The pulse sequence was velocity-compensated for in-plane flow, but through-plane flow across the ASD produced a velocity- induced phase change. Magnitude and phase images were reconstructed for each cine loop. The phase display was set to avoid extremes of windowing, looking for images in which further adjustments within a range of width and level had minimal effect on apparent defect size, consistent with a flat velocity profile with sharp “shoulders” at the orifice. The acquisition with optimal positioning at the orifice was repeated with full velocity compensation in all directions, producing a fourth cine loop in which the through-plane velocity-induced phase changes had been eliminated. In many patients, cine images were also acquired in the four-chamber view at the level that best showed the ASD in the corresponding spin-echo series.
Before the first cine acquisition, the gradient-echo pulse sequence was fine-tuned to ensure that the read gradient was balanced, resulting in a more nearly flat background when phase images were reconstructed. Because of hardware limitations, the shortest echo time achieved was 18.4 ms with full velocity compensation in a 36-cm field of view. An effective repetition time of ≈45 ms was used in combination with a steep excitation angle (60° to 90°) to maximize magnitude contrast at the ASD from flow-related enhancement in the shunt stream. Velocity encoding set to 80 to 120 cm/s gave satisfactory phase contrast at the ASD without aliasing in Signa images.
The greatest lengths over which signal was absent in the interatrial septum in the four-chamber and short-axis spin-echo images were taken as the apparent major and minor diameters of the defect. A spin-echo ASD area was calculated from the standard formula for the area of an ellipse. The cine ASD area was measured by planimetry of the circumference of the cross section of the ASD flow stream as seen in selected frames from the en face cine loops. The maximal ASD diameter by cine MRI was measured directly from the same frames. A mean of 9 (range, 2 to 27) cine images (magnitude and/or phase) were analyzed and the measurements averaged for each patient. In 3 of the 30 patients for whom correlative (surgical or catheterization) data were available, only spin-echo MRI measurements were obtained.
Fourteen patients in the study group (12 with a secundum ASD, 1 with an ostium primum ASD, and 1 with a sinus venosus ASD) had their defects closed surgically. In 12 of these patients, a cardboard template was cut during surgery to match the size and shape of the defect just before suturing was to begin. The true major diameter of each defect was measured directly from the templates, and the area of each defect was determined by planimetry of a tracing of a projected contour of each template. In 1 patient, the major and minor diameters of the defect were measured during surgery, but no template was made. In the patient with the ostium primum defect, a spin-echo image in the plane of the ASD was used as the reference standard.
In 19 patients, including 3 who later went to surgery, a defect diameter was measured in the catheterization laboratory by a balloon withdrawal technique described elsewhere.5 The shaft of the catheter proximal to the balloon was used as the magnification standard, with an external size standard as a backup reference. In several cases, measurements could also be obtained of the width of the stream of contrast medium crossing the atrial septum during filming in the cranially angulated left anterior oblique view of an injection into the right upper lobe pulmonary vein.
The balloon diameter or the angiographic diameter or an average of the two was used as the catheterization measurement of defect size, depending on the reader’s assessment of which measurement provided the clearest definition of the defect edges. From the catheterization-derived diameter, the defect area was calculated assuming a circular orifice. In the three patients who went on to surgical closure of the defect, a template was also obtained, which was used in preference to the catheterization measurement as the true standard for comparison with the MRI data.
Fig 2A⇓ shows the defect orifice as defined by the transseptal flow in a selected magnitude or phase image from a cine acquisition in the plane of the ASD for each surgical patient in whom there were matching views of the reference standard; the latter are seen in Fig 2B⇓, allowing observation of the close correspondence between the true defect shapes and the MR images.
The correlations of MRI measurements with catheterization or surgical data are shown for ASD major diameter in Figs 3A⇓ (spin echo) and 3B (cine MRI) and for defect area in Figs 3C⇓ (spin echo) and 3D (cine MRI). The regression line equations, SEEs, and correlation coefficients (r) obtained are given in the figures.
If a separate correlation is derived from the data in Fig 3B⇓ for the surgical subgroup, the r value for cine MRI maximum diameter versus template measurements (the better of the two reference standards) is r=.93. For the group measured by balloon sizing, it is r=.75.
If we exclude the two patients in whom no interatrial communication with shunt flow was demonstrated (the two points on the x axis in Fig 3B⇓), a more meaningful estimate of the accuracy of cine MRI for measuring maximum diameter is obtained, because this MRI method does not apply to these two cases. (The technique depends on transseptal flow; the absence of such flow provides no information about the size of potential defects and does not mean that the size must be zero.) The equation for the regression line for cine MRI diameter versus catheterization or surgical data then becomes y=0.78x+5.7, r=.93, and SEE=3.4 mm.
The average defect size in this study was 23 mm, as calculated both from the cine MRI data and from the surgical/catheterization data. Thus, cine MRI measurements of the largest diameter of the shunt flow stream across the septum did not systematically overestimate or underestimate the defect size. By spin echo, however, the size of secundum ASDs was larger than the reference standard by an average of 48% for major diameter and 125% for defect area; the spin-echo area was larger than the standard in all but 4 of these 28 patients. Overestimation by spin echo is also suggested by the higher intercepts of the spin-echo regression lines on the MRI axes in Fig 3⇓. The dimensions of secundum ASDs from spin-echo images were larger than those by cine MRI in all but three measurements of both major diameter and area.
Patients with a clinically suspected ASD first require confirmation of this diagnosis. Secundum ASDs may present diagnostic difficulty because of false-positive diagnoses caused by “signal dropout” commonly observed in the central portion of the interatrial septum in spin-echo images. This is related to the normal thinning of the septum at the fossa ovalis, and direct visualization of the shunt flow with a bright-blood cine MRI technique2 is frequently necessary to diagnose an ASD. This technique may reveal the defect dimension (defined by the transseptal flow) to be significantly smaller than the extent of the signal absence in the septum (Fig 4⇓). Thus, spin-echo imaging can give erroneously large measurements of ASD size and is usually by itself not adequate for evaluating secundum ASDs. Ostium primum and sinus venosus ASDs, on the other hand, are located primarily in thicker parts of the septum, and these lesions are defined more reliably by the septal discontinuity in spin-echo images.
Conventional cine MRI may also be inadequate for defining an ASD, because in some patients a definite signal void does not develop in the region of the jet (Fig 5A⇓). This is presumably related to a low interatrial pressure gradient and to the pulse sequence used. In such cases, shunt flow can be missed entirely in magnitude images. One solution to this problem is to apply a spatially selective presaturation pulse during the cine acquisition.6 Alternatively, phase-reconstructed cine images acquired with appropriate velocity encoding7 can provide greatly improved definition of the ASD stream (Fig 5B⇓).
Even when the shunt flow is adequately visualized, a cine acquisition at a single level through an ASD may not fall at its widest diameter. A stack of contiguous cine images across the defect would therefore be needed to define the widest dimension in that view and its orthogonal extent. However, this increases imaging time and/or spin saturation across slices, and the resolution in the stacked direction would be low (determined by the slice thickness). Additional cines of the jet acquired instead in a view orthogonal to the first could again fail to align with the maximal defect diameter. Cine MRI in views that show the long axis of the jet are therefore suboptimal for fully defining the size and shape of ASDs.
Our experience has been that the best strategy for defining ASD shape and dimensions is to acquire cine images of the shunt stream in a cross section located precisely in the plane of the orifice, as defined in “Methods.” The ASD is then outlined by flow-related signal enhancement in the flow crossing the imaging plane through the defect (magnitude images in Fig 2A⇑); this becomes an effective additional contrast mechanism that is not available in views in which the flow lies within the imaging plane.
Definition of the ASD in this view from flow contrast at the orifice is further amplified if the cine images are acquired with velocity encoding in the slice selection direction (ie, in the direction of ASD flow). In the resulting phase images, the shunt flow is seen as a velocity-induced phase change relative to the immediate surroundings (phase images in Fig 2A⇑). This may allow identification of an ASD that is not apparent in the magnitude image, or the phase image may define the ASD circumference more completely or more sharply. Minor signal intensity variations in a magnitude image can be windowed to appear deceptively as an ASD, and phase images should always be used to confirm whether or not an area of questionable flow-related enhancement is in fact generated by inflow through an ASD.
In this study, a localized phase change was verified as resulting from ASD shunt flow rather than another cause, by observation of its disappearance in a second acquisition at the location best positioned in the orifice, this time with full velocity compensation in the slice direction (Fig 6⇓). If these two sets of phase images are subtracted (and in the absence of eddy current effects), all background phase, including variations related to field inhomogeneity and chemical shift effects, should cancel out, leaving a phase-contrast image that shows only phase changes encoded by and proportional to velocity components orthogonal to the image plane. Although phase-velocity maps created by such difference images have a “cleaner” appearance (Fig 2A⇑, top left), we did not routinely perform the subtractions because the process was not automated on our system, and the unsubtracted phase images could still be readily interpreted and ASD dimensions measured without this extra step.
Proper selection of cine frames at the orifice is critical to the accuracy of ASD size measurements. During the cardiac cycle, the jet structure moves to a limited extent with the orifice in an axial direction in and out of the en face imaging plane (which is fixed in space). The cross-sectional images of the shunt flow selected for measurement must therefore be verified against the known morphology of the jet (Fig 7⇓). Frames acquired near end systole and during the early diastolic filling period usually demonstrate greatest flow (in agreement with echo Doppler observations8 ) and therefore show the shunt orifice most clearly, although timing at end diastole may also give good definition of the flow stream.
One might expect the defect size to be overestimated if measured in images acquired upstream of the orifice where the jet converges, as well as downstream where it diverges. In some cases, however, the apparent cross section of flow in front of the orifice was somewhat smaller than the actual orifice size. This is probably because with increasing displacement away from the central axis of flow, the blood converges along trajectories that are progressively less orthogonal to the imaging plane. Consequently, signal from this peripheral zone upstream will not appear bright because the flow-related enhancement mechanism becomes attenuated for these spins, and the reduced component of velocity along the slice selection gradient will result in a less definitive phase change in this zone.
Cine MR magnitude images in the plane of an ASD will frequently exhibit a halo of low intensity surrounding the bright signal from the coherent flow at the core of the jet. With shorter echo times, this signal void becomes less prominent; flow-related signal enhancement or the phase contrast generated by the selected velocity-encoding gradient should not be compromised, however. At the orifice (Fig 7⇓ at 0 mm), the low signal appears as a narrower rim, consistent with shear effect; this occurs within the borders of the defect, which we therefore measured to the outer edge of the signal void. Location at the orifice is confirmed by absence of such shear effect in the adjacent upstream cine image. A well-developed circumferential signal void (Fig 7⇓ at +6, +12, and +20 mm) suggests a level somewhat downstream from the orifice, where the jet has an enlarging peripheral zone of turbulence that could lead to overestimation of defect size. Such a cine acquisition should be repeated with the slice location shifted upstream as necessary to position the slice at the orifice.
The spin-echo measurements of diameter or area correlated less well with catheterization or template data than did the cine MRI measurements (r=.70 versus r=.88 for major diameter and r=.62 versus r=.90 for area). The results also show the spin-echo measurements to be significantly larger compared with both cine images and reference standards. We believe that this is because the spin-echo images may show absence of signal where there is septal thinning within the fossa ovalis adjacent to the ASD, thus increasing the apparent defect, whereas the cine images define the real ASD by a cross section of the shunt flow as it is shaped by the orifice.
Cine MRI of shunt flow at the orifice does have some important limitations. The two PFOs in this study were identified by echo but were not demonstrated by MRI. Whereas ultrasound studies permit use of hemodynamic maneuvers such as the Valsalva maneuver to induce transient transseptal flow detectable with agitated saline contrast where no such flow is present in the resting state, such maneuvers cannot be sustained during an MRI acquisition over several minutes, and significant potential defects may exist at PFOs in the absence of visible shunt flow at rest in MR cine images. This limitation of MRI may be overcome in the future by the use of Valsalva maneuvers during ultrafast MRI acquisitions. Also, if the shunt flow is not orthogonal to the cine imaging plane (the plane of the atrial septum), the flow-related enhancement and phase-contrast effects, which this technique depends on, will be compromised (Fig 8⇓).
We have used the Lock Clamshell Occluder (USCI Division, CR Bard Inc) for percutaneous ASD closure, with the requirement that the maximal defect diameter must not exceed 50% of the device diameter.5 The diameter of the largest occluder was 40 mm, and only patients whose defects measured ≤20 mm were therefore candidates for this procedure. In this study, all but 2 of the 27 patients in Fig 3B⇓ were correctly stratified by cine MRI as candidates for the clamshell versus requiring surgery (patients 1 and 2). With the measurement uncertainty for cine MRI found in this study, patients with MRI diameters from 20 to 27 mm (6 patients in Fig 3B⇓; the light dotted line is at 2 SD above the 20-mm cutoff, as shown in the “Results” above) could still be potential nonsurgical candidates. They could be offered catheterization with a view to percutaneous closure pending definitive confirmation by balloon sizing during the procedure that the defect dimension would allow this.
Other ASD closure devices currently under evaluation9 10 also typically aim for a 2:1 ratio of device to defect diameter, and a patient stratification scheme similar to the above could still be used. MRI can provide an estimate of the septal rim around a secundum defect from four-chamber and short-axis as well as from en face views. MRI can also relate the device size to the space available (width of the interatrial septum between the arrows in Fig 1B⇑). Together with knowledge of the orifice shape (Fig 2A⇑ and 2B⇑), including its minor diameter (which may be particularly useful with new “self-centering” devices), this should permit close tailoring of the device to the individual case.
Multiple two-dimensional and color-flow echocardiographic studies have reported correlation coefficients relating ASD dimensions by ultrasound to surgical or balloon measurements. These have ranged from r=.50 to r=.91 for transthoracic echo11 12 13 14 and from r=.73 to r=.92 for TEE studies.14 15 16 17 The correlation coefficient of r=.93 found in this study with phase-contrast cine MRI compares favorably with the best results reported by ultrasound, and the 3.4-mm SEE for MRI is somewhat less than the SEEs reported for echo (4.3 to 4.5 mm).12 14 15 The cine MR images provide a unique en face view, which is valuable because the complete orifice is visualized, thus defining shape as well as dimensions. A similar view of ASDs has recently also been described for three-dimensional reconstructions of the interatrial septum from TEE ultrasound data.18 The role of phase-contrast cine MRI relative to TEE for evaluation of ASDs before closure is yet to be defined. We think that both are strong techniques and that ultimately a number of centers will choose to use MRI for initial stratification to surgical versus catheter intervention because of patient comfort and convenience and the ease with which the entire defect is defined and because MRI in addition allows accurate noninvasive quantification of the shunt,19 20 21 potentially directly from the en face cine images. The TEE study could then be reserved to aid during subsequent percutaneous closure procedures.
We conclude that MRI evaluation of ASDs for nonsurgical closure allows clear localization of the defect in the interatrial septum, with definition of its type as well as its proximity to other cardiac structures. The size of a secundum defect as it appears from the signal loss in the central part of the atrial septum in spin-echo images should not be used as a measure of defect dimensions because of potentially significant overestimation. However, phase-contrast cine MR images of the cross section of the shunt stream at the orifice can yield ASD dimensions and shape with sufficient accuracy to identify patients requiring surgical repair versus those who may be suitable for nonsurgical ASD closure by newly developed catheterization techniques. Small ASDs can be sized as long as transseptal flow is present in the resting state, but cine MRI in the format described here is not a technique for evaluating PFOs; MRI methods may be extended to this application in the future. Careful attention to the multiple details of imaging technique, particularly with respect to localization of the cine imaging plane at the orifice, is required to obtain studies with good image quality from which quantitatively accurate measurements can be made.
Selected Abbreviations and Acronyms
|ASD||=||atrial septal defect|
|MRI||=||magnetic resonance imaging|
|PFO||=||patent foramen ovale|
This study was supported in part by a Canadian Heart Foundation fellowship grant. We would like to thank Drs M.E. King, P. Lang, R.S. Lees, R.A. Levine, A.J. Marelli, S.M. Nidorf, and J.D. Thomas for thoughtful discussions of the subject. We would also like to thank Dr H.E. Waldman for help with the balloon sizing and Drs C.W. Akins, M.J. Buckley, and J.C. Madsen for help with obtaining the surgical data. We are grateful to Joe Malysa and the Technicare/Plexar staff for providing the basic version of the gradient-echo pulse sequence source code along with helpful information. Finally, we thank our patients for their participation in this imaging study.
Presented in part at the 40th Annual Scientific Session, American College of Cardiology, Atlanta, Ga, March 3-7, 1991, and published in abstract form in J Am Coll Cardiol (1991;17:326A).
- Received February 27, 1995.
- Revision received July 31, 1995.
- Accepted August 6, 1995.
- Copyright © 1995 by American Heart Association
Theissen P, Sechtem U, Mennicken U, Hilger HH, Schicha H. Non-invasive diagnosis of atrial septal defects and anomalous pulmonary venous return by magnetic resonance imaging. Nuklear Medizin. 1989;28:172-180.
Bridges ND, Hellenbrand W, Latson L, Filiano J, Newburger JW, Lock JE. Transcatheter closure of patent foramen ovale after presumed paradoxical embolism. Circulation. 1992;86:1902-1908.
Hartnell G, Sassower M, Finn P. Selective presaturation increases the accuracy of magnetic resonance angiography in the detection of intracardiac shunts. Circulation. 1993;88(suppl I):I-83. Abstract.
Underwood SR, Firmin DN, Klipstein RH, Rees RSO, Longmore DB. Magnetic resonance velocity mapping: clinical application of a new technique. Br Heart J. 1987;57:404-412.
Das GS, Voss G, Jarvis G, Wyche K, Gunther R, Wilson RF. Experimental atrial septal defect closure with a new, transcatheter, self-centering device. Circulation. 1993;88(pt 1):1754-1764.
Hanrath P, Schlüter M, Langenstein BA, Polster J, Engel S, Kremer P, Krebber HJ. Detection of ostium secundum atrial septal defects by transesophageal cross-sectional echocardiography. Br Heart J. 1983;49:350-358.
Holmvang G, Dinsmore RE, Edelman R, Thompson RC, Miller SW, Leavitt M, Brady TJ, Kantor HL. Quantitation of central circulatory shunts by NMR. Circulation. 1986;74(suppl II):II-5. Abstract.
Hundley WG, Li HF, Lange RA, Pfeifer DP, Meshack BM, Willard JE, Landau C, Willett D, Hillis LD, Peshock RM. Assessment of left-to-right intracardiac shunting by velocity-encoded, phase-difference magnetic resonance imaging: a comparison with oximetric and indicator dilution techniques. Circulation. 1995;91:2955-2960.