a) Artifacts associated with instrumentation errors.

• Flood field non-uniformity: This may be secondary to defective photomultiplier tubes, damaged collimators, problems with camera electronics, or defects in the sodium-iodine crystal. To detect such abnormalities, intrinsic and extrinsic flood fields should be obtained as described in protocols. Flood field non-uniformities can be detected either visually or quantitatively by appropriate QC software or manual calculations. When a SPECT study is acquired using a non-uniform camera, each flood field defect will be propagated into a circular arch generating ring artifacts that can potentially cause irregularities in myocardial count density. These artifacts are more evident towards the center of the reconstruction matrix and can be either "hot" or "cold". With dual-head cameras, minor differences in detector sensitivity can also cause image artifacts even if absolute values are acceptable separately.
• Center-of-rotation (COR) error. The COR allows SPECT data to be back-projected onto a central point in the volumetric matrix. Exact centering of data reconstruction is critical to accurately reproduce anatomic structures and count density distribution. Error in the COR will result in misregistration of tomographic data, image blurring and artifactual SPECT myocardial perfusion defects. The myocardial image may appear to be misaligned or skewed specially towards the apex, which is more evident in the horizontal long axis plane. With dual-head, variable angle cameras COR alignment can potentially represent a more complex problem.
• Camera head tilt. Similar to the problem with COR, if one detector is tilted while using a variable angle multihead camera, the detector heads will not record exactly the same image. Back projection and reconstruction of these different images into a single cardiac image may also produce regional myocardial perfusion artifacts.
• Detector-to-patient distance. Although not technically an artifact per se, minimizing detector-to-patient distance is a very important technical consideration to assure optimal image quality and diagnostic accuracy of the procedure. Increased distance results in suboptimal defect contrast and spatial resolution. When defining the camera orbit, radius of rotation should be kept as smaller as possible. Also, with non-circular orbits, careful must be taken to avoid significant variations in radius of rotation at different angles since backprojection of images of different resolution can also produce artifacts.

b) Patient related artifacts.

• Patient motion. This is one of the most common sources of SPECT myocardial perfusion artifacts. During SPECT acquisition, patients may move in the vertical (axial), lateral or rotational direction. The best means to detect patient motion is to observe the planar raw images of the SPECT acquisition in cine mode, but condensed images such as sinogram and linogram are also useful. Whether patient movement is abrupt or gradual, myocardial perfusion image artifacts may be created. The cause of the reconstruction artifact is similar to that due to an erroneous COR, but with patient motion the heart itself is at a different location during different portions of the SPECT acquisition. During reconstruction, data are backprojected onto different points of the volumetric matrix, and thereby a misregistration error occurs.
  

Respiratory motion is very prominent in some patients but the images are often still of diagnostic value, because there is a periodical movement of the heart in the axial direction causing some blurring of the backprojected image but usually not simulating a perfusion defect. A change in body position at the middle of the acquisition is much worse in consequence. Characteristically, with patient motion reconstructed images demonstrate opposed defects in contralateral walls. Frequently the image contains "tails" streaming from the edges of the defect. The overall visual effect is known as the "hurricane sign" in the short axis views, while in the horizontal long axis misalignment of the lateral and septal wall towards the apex can be apparent as described for COR errors. Motion correction can be attempted although most software approaches are imperfect and sometimes represent additional sources of error by themselves.

The technologist can be of great help in minimizing patient motion. SPECT acquisition should be fully explained to the patient and he/she should be encouraged not to talk, yawn, sigh or fall sleep (and snore). The lumbar curvature of the back should be supported, the knees slightly elevated and supported in order to minimize back strain, and the left arm and shoulder supported and restrained. Although SPECT can be performed with both arms at the side, in order to minimize detector distance and avoid the attenuation caused by the arms, elevation of at least the left arm above the head is preferable. Commercially available arm holders are useful for this purpose. Multiheaded detector systems have been beneficial to decrease SPECT image acquisition time, improve patient tolerance, and thereby decrease the possibility of patient motion. On the other hand, a single motion episode will be duplicated with a dual-head camera since it will be "seen" simultaneously by both detectors at different angles.

If significant motion is detected and reconstruction artifacts are present, a new acquisition is recommended. However, if repeating a stress ²⁰¹Tl study significant redistribution could have occurred and scheduling the patient for a different day may be preferable. This is not a problem with ^99mTc-labelled agents.

• Soft tissue attenuation. The effect of soft tissue attenuation may be generalized or localized. Generalized attenuation, encountered in obese patients and in individuals with a large chest circumference, results in decreased counts density and poor image quality. More frequent are fixed, localized soft tissue attenuation artifacts secondary to the left hemidiaphragm, large abdomen, the left breast, and lateral chest wall fat, usually mimic myocardial scarring. Breast attenuation generates apparent hipoperfusion of the anterior wall and can be seen not only in women with large breasts but also in those with relatively small but dense breasts. Shifting breast attenuation can be caused by different position of the breast in the stress and rest acquisitions. Such differences will result in attenuation artifacts that affect different portions of the left ventricle in the stress and rest images and thereby simulate stress-induced ischemia and/or reverse distribution. Localized diaphragmatic attenuation generally creates a fixed inferior defect. Unlike the breast, the position of the left hemidiaphragm is relatively stable and therefore, shifting attenuation artifacts are relatively rare.

Repeating the acquisition with the left breast elevated and restrained with tape, or in the prone position, can diminish the attenuation effect of breast tissue and the left hemidiaphragm, respectively. Attenuation correction, now commercially available on many scintillation camera systems, is of benefit in minimizing or eliminating many of these artifacts. However, the interpreting physician must be cautions of unique artifacts produced by current attenuation correction algorithms, including truncation, defects produced by inadequate transmission source count density and uniformity, exaggeration of apical thinning, and scatter of subdiaphragmatic activity into the inferior wall of the left ventricle. Gated perfusion SPECT is an efficient alternative to minimize false positive studies because portions of the heart with attenuation will have preserved wall motion and thickening, while true fybrotic tissue will not.

c) Image processing and display artifacts.

• Filtering. Commercial manufacturers have incorporated empirically determined optimal filters for SPECT myocardial perfusion protocols. Technologists and interpreting physicians should be aware of the consequences of altering filter parameters and they should also be able to recognize the effects of applying standard filters to images of unusually low or high count density.
  Decreasing the filter cutoff (critical frequency) will render very smooth images with loss of detail and contrast resolution, and hence with decreased sensitivity in detecting perfusion abnormalities. On the other hand, increasing the cutoff value will accentuate high frequency data and exaggerate noise, giving the tomogram an unpleasant, difficult to interpret appearance. Likewise, lowering the critical frequency will decrease the apparent extent and severity of a true perfusion abnormality, whereas increasing the critical frequency may appear to accentuate the defect. The result is similar if the filter cutoff is held constant and the count density of an image decreases: true perfusion abnormalities as well as attenuation artifacts may appear more evident and severe. It is possible to adjust a filter to match the count density of an image (adaptive filtering). However, unless the technologist is exceptionally knowledgeable about SPECT filters, it is probably advisable not to adjust prescribed filters but to instead anticipate filtering effects in particular patients.
  
• Adjacent subdiaphragmatic activity. Abdominal tracer concentration adjacent to the inferior wall of the left ventricle may create significant image artifacts. ²⁰¹Tl localizes in the liver, and MIBI and tetrofosmin have initial liver uptake and are then excreted via the hepatobiliary tract to the duodenum. The tracer moves distally in the small bowel, which may be positioned in the left upper quadrant of the abdomen, or it may sometimes reflux into the stomach. Subdiaphragmatic activity may artifactually increase inferior wall count density due to scatter, or decrease count density secondary to the application of the ramp filter during backprojection process.

Scatter: photons emanating from subdiaphragmatic activity may artifactually increase inferior wall count density. This phenomenon may obscure true inferior wall perfusion defects or, in normal patients, the inferior wall may become relatively hot. By subsequent image normalization an artifactual defect may be created in the contralateral anterior wall. Such artifacts are more common with ²⁰¹Tl than with 99mTc because of the larger Compton angle, and when an all purpose collimator is used instead of a high resolution collimator.

Ramp filter artifact: with filtered backprojection, the ramp filter is used to eliminate the star artifact associated with reconstruction by applying negative values adjacent to the projection profiles. When there is intense, localized tracer concentration next to the heart, present in the liver, stomach, or small bowel, the negative values of the corresponding projection profiles may cause an artifactual decrease in count density in the inferior wall of the left ventricle. One way to check if the finding is an artifact caused by this effect is to raise the cutoff frequency of the filter, which will result in a noisier image but the "cold" area will be reduced. This effect is not evident with iterative reconstruction algorithms.

• Improper selection of the apex and base for polar map reconstruction. Accurate and reproducible selection of the apex and base of the left ventricular myocardium is necessary in stress and rest images. Positioning the limits for slice selection too far basally will result in an apparent basal myocardial perfusion defect, while positioning the limits distal to the apex will result in an apparent apical defect. In contrast, positioning slice limits too tightly, so that they do no encompass the entire heart, may result in underestimation of the size and extent of a defect.
  
• Errors in axis reorientation. If the long axis of the ventricle is defined incorrectly for reorientation on either transaxial or midventricular vertical long axis slices, the geometry of the heart in subsequently reconstructed orthogonal tomographic slices can be distorted. Consequently, the apparent regional count density can be altered resulting in artifactual perfusion defects.
  
• Inadequate image display. Stress and perfusion tomograms should each be individually normalized to maximal myocardial count density so that stress and rest perfusion defects can be compared and assessed for reversibility. If images are incorrectly normalized to non-cardiac activity, particularly subdiaphragmatic tracer concentration, accurate comparison of stress and rest images is not possible.

d) Artifacts related to non-coronary disease.

• Left bundle-branch block (LBBB). This conduction abnormality produces apparent septal hypoperfusion that can be either reversible or fixed, lowering the specificity of the study for detection of ischemic heart disease. In gated studies, however, the septal wall usually exhibits paradoxical motion but with preserved thickening, aiding in the differential diagnosis.
  
• Left ventricular hypertrophy. It can cause localized hot spots with relative hypoactivity of remote areas, and other myocardial non-homogeneities.
  
• Long membranous /short muscular septum. It can produce false impression of hypoperfused basal septum.
  
• Apical thinning. This can be either a normal physiological variant or an artifact produced by varying spatial resolution of the acquired images (higher at the apex because of detector proximity). It usually simulates a fixed defect and can be differentiated from scar by gating since myocardial thickening and wall motion will be preserved.
  
• The eleven / seven o'clock defect. More frequent at anterior-septal region (11 o'clock) in short-axis tomograms, this pseudo-defect probably represents attenuation by the right ventricle since it usually lies, together with the 7 o'clock defect, close to the insertion points of the right ventricular wall. Again, gated SPECT can aid in diagnosis.

e) Gating artifacts.

• With severe arrhythmia or heart rate changes during SPECT acquisition, short or long cardiac cycles will be rejected according to the tolerance window selected as mentioned in the section on instrumentation and acquisition parameters. If the proportion of irregular beat rejection varies during acquisition, the total number of cardiac cycles acquired for each projection will be different, assuming that each projection image is acquired for the same length of time as is the standard protocol for gated perfusion SPECT. Therefore, projection images will differ in count density. When viewed in endless loop cine mode, the projection images will appear to flash. Acquisition of a variable number of counts in the projection images may potentially result in errors in filtered backprojection and consequently perfusion artifacts. However, it has been reported that clinically significant perfusion artifacts are produced only with severe arrhythmias, such as associated with atrial fibrillation.
  
• Variation in cycle length also can produce low count density images specially those from the final portion of the cardiac cycle (usually frames 6-8). As a result of the normalization process, these images are assigned an average count rate after multiplying by a certain factor, with the effect of increasing noise value which is already present in excess. During backprojection, this noise significantly alters the quality of reconstructed images, potentially introducing hot and cold artifacts of variable shape and location.
  
• Gating with waves other than R wave can result in false interpretation of end diastole, with a volume curve starting elsewhere in the cardiac cycle. This fact, however, rarely causes a problem other than a volume curve of unexpected shape.
  
• The ECG signal can be lost during acquisition because of patient motion or electronic problems, thus causing severe deterioration of functional data. The technologist must constantly supervise the gated acquisition process, preferably with the aid of a ECG monitor or an acoustic signal.