Chapter 5 Single-Photon Emission Computed Tomography Artifacts

INTRODUCTION

Single-photon emission computed tomographic (SPECT) imaging of myocardial perfusion provides a sensitive means of detecting and localizing coronary artery disease (CAD) and assessing left ventricular (LV) function. However, the method suffers from the potential for poor specificity due to image artifacts related to both patient and technical factors. By recognizing the sources of such artifacts and understanding the means available to avoid them, the technologist and interpreting physician can substantially improve the specificity of SPECT studies and more meaningfully contribute to appropriate patient treatment.

TECHNICAL ARTIFACTS

Flood Field Nonuniformity

Flood field nonuniformity will result in “ring” artifacts in reconstructed SPECT images. These relatively photon-deficient rings may be apparent in tomographic slices and in severe cases may also appear in polar coordinate maps (Fig. 5-1). Because patients are frequently positioned differently within the camera field for rest and stress scans, flood field artifacts may occur in differents of the myocardium, mimicking reversible or partially reversible defects. Therefore, routine acquisition and inspection of intrinsic and extrinsic flood fields acquired according to vendor recommendations are absolutely critical to avoid such artifacts. Flood fields are routinely acquired the first thing in the morning each working day. However, if ring artifacts appear in clinical SPECT images, it may be necessary to reacquire flood field images in the middle of the day.

Figure 5-1 Flood field nonuniformity resulting in ring artifacts and artifactual myocardial perfusion defects.

(Reproduced with permission from Iskandrian AE, Verani MS [eds]: Nuclear Cardiac Imaging: Principles and Applications, 3rd ed. New York: Oxford University Press, 2003.)

Center of Rotation and Camera-Held Alignment Errors

If the camera center of rotation (COR) is incorrect, filtered backprojection during SPECT reconstruction will result in image misregistration and apparent misalignment of the myocardial walls (Fig. 5-2). Technically, this is similar to the error created by cardiac motion. However, unlike motion artifacts, those due to COR error are usually more systematic and predictable. The severity of the apparent defect is directly proportional to the magnitude of the COR error. An error similar to that produced by the wrong center of rotation is produced when the detector is not aligned perpendicular to the radius of rotation.

Figure 5-2 Schematic representation of camera center of rotation error and resultant artifactual myocardial perfusion defects. Defects are in opposite sides of the “heart,” and “tails” of activity are present extending from the edges of the defects.

(Reproduced with permission from Iskandrian AE, Verani MS [eds]: Nuclear Cardiac Imaging: Principles and Applications, 3rd ed. New York: Oxford University Press, 2003.)

To avoid these artifacts, camera COR checks should be performed according to the camera manufacturer’s guidelines. The resultant plots should be inspected regularly by a knowledgeable technologist and physician before the camera is used for patient studies. With the observation of a myocardial defect suspected of representing such an artifact, the COR determination should be repeated immediately. Likewise for variable-angle multiheaded detector systems, camera head alignment should be assessed periodically, particularly after camera repair and whenever the detectors are impacted by objects such as stretchers.

Errors in Selecting Oblique Cardiac Axes and Subsequent Polar Map Reconstruction

If the long axis of the left ventricle (LV) is defined incorrectly on either the transaxial or midventricular vertical long-axis slice, the geometry of the heart in subsequently reconstructed orthogonal tomographic slices can be distorted. Consequently, the apparent regional count density can be altered in polar maps, resulting in apparent perfusion defects. These may be accentuated by quantitative analyses in which patient data are compared to normal files (Fig. 5-3). In polar map reconstruction, such errors most often occur in basal myocardial regions at the periphery of the bull’s-eye plot, owing to foreshortening of one of the myocardial walls (Fig. 5-4). Also, the apex, which often demonstrates physiologic thinning and decreased count density, is displaced from the exact center of the polar plot. The displaced apex may also consequently result in an artifact. Misregistration of perfusion defects in short-axis slices and polar maps is not infrequently encountered in patients with sizable severe infarcts.

Figure 5-3 If the long axis of the left ventricle is selected inconsistently, in stress and rest transaxial tomographic slices, perfusion defect localization, cavity configuration, and the size of the basal septal defect due to the membranous septum will be represented inconsistently in reconstructed short-axis tomograms.

(Reproduced with permission from Iskandrian AE, Verani MS [eds]: Nuclear Cardiac Imaging: Principles and Applications, 3rd ed. New York: Oxford University Press, 2003.)

Figure 5-4 Stress tomograms (top row) were reconstructed correctly in this normal patient, whereas the long axis of the left ventricle was selected incorrectly for the resting tomograms. As a result, the ventricular cavity in the resting short-axis slices appears elliptical, and there is an artifactual perfusion defect in the basal inferoseptal region.

(Reproduced with permission from DePuey EG, Garcia EG, Berman DS [eds]: Cardiac SPECT Imaging, 2nd ed. New York: Raven, 2001, p 195.)

Selection of Apex and Base for Polar Map Reconstruction

Accurate and reproducible selection of the apex and base of the LV myocardium is necessary in stress and rest images. Positioning limits for slice selection that lie too far apically or basally will result in apparent perfusion defects (Fig. 5-5). In contrast, positioning slice limits too tightly, so that they do not encompass the entire heart, will potentially result in underestimation of the size and extent of a defect. Likewise, in regions of normal myocardium, maximal myocardial count density may not be correctly sampled.

Figure 5-5 When the apical limits for reconstruction are defined distal to the left ventricular myocardium in stress images (A), an artifactual extensive and severe apical ischemic defect is produced in the polar plots (B).

(A, Reproduced with permission from DePuey EG, Garcia EG, Berman DS [eds]: Cardiac SPECT Imaging, 2nd ed. New York: Raven, 2001, p 98. B, Reproduced with permission from Iskandrian AE, Verani MS [eds]: Nuclear Cardiac Imaging: Principles and Applications, 3rd ed. New York: Oxford University Press, 2003.)

Arrhythmias and Gating Errors

If arrhythmia is present or if the heart rate changes during SPECT acquisition, inordinately short or long cardiac cycles will be rejected during an 8- or 16-frame/gated acquisition. If the degree of the regular beat rejection varies during the SPECT acquisition, the number of cardiac cycles acquired for each projection image may vary if each projection image is acquired for the same length of time. Therefore, projection images will vary in count density. When viewed in endless loop cinematic format, the projection images will appear to “flash.” The most serious effect of arrhythmia is a decrease in image count density due to these “discarded” cardiac cycles. However, only with severe arrhythmias associated with atrial fibrillation and frequent premature ventricular contractions are clinically significant perfusion artifacts encountered. Although this problem may be overcome by prolonging SPECT acquisition until a prescribed number of regular cardiac cycles have been acquired, such longer acquisition times increase the probability of patient motion due to discomfort or anxiety. Recently, some manufacturers have established a “9th bin” wherein rejected cardiac cycles are stored temporarily and excluded from reconstruction of the gated tomograms, but added back into the summed, nongated images. Alternately, other manufacturers allow simultaneous acquisition of gated and nongated images so that arrhythmic beat rejection will not result in low count-density ungated images.

If the accepted cardiac cycles vary in length, and because each cycle begins at the R-wave used for gating, relative shorter beats will not fill the entire 8 or 16 “bins” of the cardiac cycle. When all of the acquired cardiac cycles are summed to create an 8- or 16-frame gated image, this will result in data “dropout” in the last few “bins” of the cardiac cycle.

Displays of cardiac image count density from individual planar projection images is now possible using some commercially available software programs (Fig. 5-6). Errors in LV volume and ejection fraction may result from gating errors, but these are beyond the scope of this chapter.¹

Figure 5-6 A, Count density is plotted on the y axis for each of the eight “bins” of an 8-frame per cardiac cycle gated post-stress myocardial perfusion scan for the 64 camera stops in the 180-degree SPECT imaging arc of a 90-degree-angled, two-detector scintillation camera. In this patient with no arrhythmia and a very regular R-R interval, the curves are superimposed for each of the eight gated “bins.” B, In this arrhythmic patient, cardiac cycles are rejected during camera stops ~8 to 24 and simultaneously during stops ~40 to 56 for the other scintillation detector. This is manifested as a decrease in counts acquired/accepted at those stops. C, In this patient, all cardiac cycles were accepted during the SPECT acquisition. However, owing to variation of the R-R interval of the accepted beats, there is a decrease in count density in the eighth gated “bin.”

PATIENT-RELATED ARTIFACTS

Soft-Tissue Attenuation (See Chapters 6 and 7)

The location of the attenuation artifact depends on the position of the soft-tissue attenuator in relation to the left ventricle. The severity of the artifact depends on the size and density of the attenuator in relation to adjacent tissue. Within the resultant SPECT image, the artifact may appear as a fixed or reversible defect or may mimic “reverse redistribution,” depending on whether the attenuator is in a constant or variable position in stress and delayed image acquisitions. The severity of the attenuation artifact also depends on the energy of the incident photon. Attenuation artifacts for technetium (^99mTc)-labeled myocardial perfusion agents will be somewhat less marked than for thallium (²⁰¹Tl).³ However, for either isotope, evaluation of regional wall motion is helpful in differentiating attenuation artifacts from myocardial scarring as a source of fixed perfusion defects.⁴^–⁶

Breast Attenuation

Because breasts vary in size, position, configuration, and density, breast attenuation artifacts are extremely variable in appearance. In addressing the characteristics of breast attenuation, it is always necessary to consider the position and configuration of the breasts with the patient in the supine position, because patients are imaged in this position with most commercially available SPECT systems. In women of average body habitus, the left breast overlies the anterolateral wall of the heart. In women with large, pendulous breasts, the breasts lie adjacent to the lateral chest wall and more often result in a lateral attenuation artifact. In some women with very large, pendulous breasts, the attenuation artifact may create apparent inferior or inferolateral perfusion defects. In women with very large breasts, the breast tissue may overlie the entire LV. In this instance, the resulting attenuation artifact may be diffuse and less discrete or may primarily involve the apex. In addition to breast position, the caudal angulation of the heart within the thorax will affect the appearance of the attenuation artifact.²^–³ Thus, in summary, the severity of breast attenuation artifacts is not necessarily directly proportional to breast size or chest circumference but may vary considerably according to the position, configuration, and density of the breast, body habitus of the patient, and orientation of the heart within the thorax. Additionally, when women are imaged upright or in a semi-upright, “reclining” position, the breasts are usually more pendulous than in the supine position. Therefore, apical and inferior breast attenuation artifacts are more frequent when women are imaged in these positions.

In some laboratories, a binder is used to flatten the breasts against the chest wall. Although this technique can be quite useful to decrease the thickness of the breast, there is often a degree of uncertainty about the exact position of the breast beneath the binder. Moreover, precise repositioning of the breast beneath the binder in stress and rest studies is difficult. Therefore, in most laboratories, breast binders are not used. For similar reasons, taping the breast upward in an attempt to avoid its overlying the heart is associated with uncertainty in reproducing the breast position for stress and rest SPECT acquisitions.

The interpreting physician must also be cognizant of factors that may further alter the position, configuration, and density of the breast. A bra positions the breast anteriorly and increases the amount of soft-tissue attenuation by thickening the breast. For this reason, it is recommended that SPECT imaging be performed with the bra off. Breast implants may be denser than normal breast tissue and may accentuate attenuation artifacts. Therefore, women who are candidates for SPECT should always be questioned about a history of breast implants or augmentation. Patients with previous left mastectomies may wear a breast prosthesis. Such prostheses may be quite dense, and women who have them must always be instructed to remove them before myocardial perfusion imaging.

Quantitative analysis of SPECT myocardial perfusion imaging is helpful, particularly to the novice, in differentiating normal, physiologic variations in radiotracer distribution from true perfusion defects. However, gender-matched normal files are derived from an average population, and normal limits are established on the basis of variations in regional count density within this population. However, this type of analysis does not take into account the marked variability in body habitus observed in the patient population referred for myocardial perfusion imaging. Therefore, quantitative analysis, instead of aiding the physician in identifying soft-tissue attenuation artifacts, may actually decrease the specificity of SPECT testing by incorrectly identifying such artifacts as abnormalities.

Attenuation correction for SPECT is now commercially available but is still a topic of intense interest and research. The methods for accomplishing it have used either scatter correction or attenuation correction using a separate radionuclide or x-ray transmission image. This approach has been demonstrated to significantly reduce breast and other attenuation artifacts in phantom and patient studies.⁷^–¹⁸ However, whereas attenuation correction may minimize attenuation artifacts created by overlying breast tissue, breast artifacts are frequently not totally eliminated.

Certain breast attenuation artifacts are unique to quantitative analysis in cardiac SPECT. The most common is an apparent inferior perfusion defect incorrectly identified by quantitative analysis in women who have had a left mastectomy. After removal of the breast, there is little or no anterior soft-tissue attenuation. In such cases, the anterior and inferior myocardial count densities are nearly the same as they are for males. Since the normal female file anticipates that the inferior wall will be more intense than the anterior wall, and in the patient study the inferior wall has an identical or lower intensity than the anterior wall, the inferior wall will be identified as abnormal (Fig. 5-7). This quantitative error can be circumvented by comparison of the patient’s data to the normal male file instead. Similar inferior artifacts have been observed in quantitative plots in women with very small breasts, and it might be argued that data from such patients should be compared to normal male limits. However, small breasts may be quite dense, and the assumption that they do not produce photon attenuation may be incorrect. Therefore, in most laboratories, the normal male file is used routinely only for women who have undergone left mastectomy.

Figure 5-7 Technetium-99m-sestamibi stress and rest tomograms (A) and quantitative polar plots (B) in a patient with a left mastectomy. Patient data are compared to a normal female file, and quantitatively an inferior defect is identified artifactually. However, when patient data are compared to a normal male file (C), which anticipates no breast attenuation, quantitative analysis demonstrates no blackened pixels in the inferior wall.

(Reproduced with permission from Iskandrian AE, Verani MS [eds]: Nuclear Cardiac Imaging: Principles and Applications, 3rd ed. New York: Oxford University Press, 2003.)

The position of the breast may vary from stress to rest if a woman wears different clothing at the time of the two SPECT acquisitions. Breast position varies considerably with a bra on and off. The position of the breast may also vary depending upon the degree of elevation of the left arm, which is preferably positioned above the head for a SPECT acquisition. A shifting breast attenuation artifact can mimic a reversible perfusion defect (i.e., ischemia). For example, if during stress SPECT acquisition, the breast lies high over the anterior chest wall, an anterior attenuation artifact may result. However, if during the resting acquisition, the breast lies more laterally and inferiorly, the artifact will involve the inferolateral wall of the LV. The resulting attenuation artifacts are therefore different in the stress and rest images, so artifactual defect reversibility as well as artifactual “reverse distribution” may result. In this example, the anterior defect will appear to be reversible, whereas there will appear to be “reverse distribution” in the inferolateral wall (Fig. 5-8).