Instrument- and computer-related problems and artifacts in nuclear medicine

Instrument- and Computer-Related Problems and Artifacts in N u c l e a r M e d i c i n e Michael K. O'Connor Advances in y-camera design over the last 5 to 10 years have improved all aspects of image quality, particularly for tomographic imaging. As system complexity increases, it becomes more important that both the technologist and physician be able to recognize the various types of artifacts that can occur in y-camera systems and their potential impact on clinical studies. A thorough evaluation of the system at installation and a comprehensive quality control program will detect most problems that can occur. The most sensitive indicator of -/-camera performance is uniformity. Because this measurement is performed on a daily basis, it is the principle tool in evaluating the status of the -/-camera. Most artifacts related to the integrity of the detector head, computer system, and hard copy device can be detected on the uniformity image. For

tomographic imaging, a quantitative determination of uniformity is needed to ensure that the system will not introduce ring artifacts into the patient data. Unlike planar imaging, the complex nature of tomographic imaging requires additional checks on system performance to ensure that other components of the system, such as the Collimator, gantry, and imaging table, are not a source of artifact in the reconstructed image data. Failure of a system component can occur at any time and may be subtle and difficult to recognize in modern systems. In such an environment it becomes imperative that any unexpected finding in a clinical study be questioned with respect to a possible malfunction of some aspect of data acquisition and analysis.

VER T H E LAST 10 years, the -/-camera has evolved from a relatively simple analog device into a complex device with a myriad of digital correction techniques designed to improve all aspects of system performance. Each additional correction or manipulation of the raw data brings with it the possibility of error. Although artifacts resulting from an error in the acquisition of the raw data may be simple to recognize, those resulting from an error in the application of the nth correction to the data may not be so obvious. The more the data are manipulated, the more difficult it becomes to distinguish what is real from what is artifactual. Therefore, it is imperative that adequate quality control be performed and that standardized imaging protocols be established to ensure reliable and reproducible clinical results. The importance of patient preparation and procedural details is discussed elsewhere in this issue. The recommended routine quality control procedures for -/-cameras have remained essentially unchanged over the last 10 to 15 years, despite the increased complexity of the modern systems. Routine quality-control procedures usu-

ally include measurement of system uniformity and resolution, 1 with the additional assessment of system center of rotation for tomographic imaging. Although such procedures are still essential, and indeed adequate for many `/-cameras in clinical use today, additional procedures may be n e e d e d to ensure accurate function of many of the newer features now present on modern systems, particularly in the area of tomographic imaging. Artifacts can arise from a number of sources: the radiopharmaceutical, the `/-camera, the patient, or the procedure. This review focuses specifically on those artifacts originating from the -/-camera and the computer system and will present some of the more common artifacts seen with -/-cameras, as well as those of a more complex nature associated with tomographic imaging. Some simple guidelines and qualitycontrol checks that may be useful in the monitoring of equipment for such problems are presented. This review is divided into two p a r t s - planar and tomographic--because some artifacts that have minimal impact on the quality of planar studies can seriously compromise the quality of tomographic studies.

From the Section of Nuclear Medicine, Mayo Clinic, Rochester, MN. Address reprint requests to Michael K. O'Connor, PhD, Section of Nuclear Medicine, Charlton 2N-213, Mayo Clinic, Rochester, MN 55905. Copyright 9 1996 by W.B. Saunders Company 0001-2998/96/2604-000355.00/0

PLANAR IMAGING

O

256

Copyright 9 1996 by W.B, Saunders Company

Uniformity The most sensitive indicator of the performance of a `/-camera system is measurement of its response to a uniform (flood) field of radia-

Seminars in Nuclear Medicine, Vol XXVI, No 4 (October), 1996: pp 256-277

ARTIFACTS

IN N U C L E A R M E D I C I N E

257

tion. 2 This measurement can be performed either intrinsically (ie, without a collimator) or extrinsically. It is the single most important determinant of ~/-camera function because most instrumentation problems adversely affect the uniformity Of the flood image. Because measurement of system uniformity usually is performed on a daily basis, artifacts resulting from system malfunction often are first apparent on images of a uniform field of radiation. Artifacts in the flood images can arise from two sources--the -/-camera and the radiation sources used to produce the flood images. For extrinsic measurement of uniformity, either cobalt 57 (57Co) or Technetium 99m (99mTc) sheet sources are used. Both of these may inadvertently introduce artifacts into the flood images. SVCosheet sources usually contain small amounts o f 56C0 and 58C0.3 These radionuclidic contaminants have a shorter half-life (70 to 80 days) than 57C0 but emit high-energy ~/-rays ( > 500 keV). Figure 1 shows the energy spectra of a recently manufactured 57C0 sheet source (1 month from date of manufacture) and a 1-yearold 57C0 sheet source. When a low-energy collimator is used, a significant amount of highenergy contamination is evident in the spectrum of the i-month-old source. During the first few months of u s e , this contamination may adversely affect the performance of the ~-camera 4 unless measurements of uniformity are performed using a medium- or high-energy collimator. Some ~/-cameras are particularly sensitive 100 - ..... ....

90 8O

LEHR Coil/,C0-57 1 month MED CoU / Co-57 1 month LEHR CoU / Co-57 1 year

70 60

50

8

40 30 20 10

9.,:.:.::.:_:_.-.-;-:.~ 50

70

90 Gamma

110

130

150

170

190

Ray Energy(keV)

Fig 1. Energy spectra acquired extrinsically using either a new (1-month-old) or old ( 1 - y e a r - o l d ) STCo sheet source. The ~,-camera was equipped with either a low-energy, highresolution (LEHR) or a medium-energy (MED) collimator. Note the varying degrees of high-energy contamination above 1 2 2 keV.

A

B

Fig 2. Thirty-million count flood images acquired using a 1-month-old S7Co sheet source on a ~/-camera equipped with ( A ) a low-energy, high-resolution collimator and (B) a mediumenergy collimator.

to high-energy contaminants and may show nonuniformities in the flood images when quality control is performed using a low-energy collimator. Figure 2 shows examples of flood images acquired using a 1-month-old 57C0 sheet on a ~,-camera equipped with a low-energy, high-resolution collimator and a mediumenergy collimator. The use of a medium-energy collimator in place of a low-energy collimator has significantly reduced the impact of these contaminants on image quality. 99mTc sheet sources usually are produced by adding several millicuries of 99mTc to a liquidfilled plastic sheet source. These sources are prone to a number of problems, including distortion of the sheet source, 5 air bubbles inside the source, poor mixing of the isotope within the source, and clumping or adhesion of the isotope to the walls of the plastic container. The last problem is often the result of using 99mTcbound to colloidal particles (eg, 99mTc-sulfur colloid) or in some chemical form that facilitates aggregation of the 99mZc. Figure 3 shows examples of artifactual nonuniformities in flood images from a variety of problems associated with refillable sheet sources. All of these problems can be avoided by careful p r e p a r a t i o n of the sheet source and routine inspection of the source to check for warping, leakage, or distortion of the plastic. Intrinsic measurement of -/-camera uniformity is inherently less prone to error than extrinsic measurement. Nevertheless, care is still needed in the set-up for this measurement. Figure 4 shows an example of a flood image

258

A

MICHAEL K. O'CONNOR

B

C

Fig 3. Flood images acquired with refillable sheet sources containing 99mTc. Image artifacts are a result of (A) hot spots caused by clumping of the 99mTc-sulfur colloid particles, (B) air bubbles within the source, and (C) poor mixing of the 99rnTc"

acquired intrinsically using a 0.1-mL point source. The 99mTc was drawn up in a 1-mL syringe and orientated with the needle pointing toward the detector. The hot spot and surrounding cold circular pattern (arrow) seen on the flood image are the result of attenuation from the walls of the needle, with the needle effectively acting as a collimator. Hence, for both extrinsic and intrinsic measurements of uniformity, care is needed to ensure that the radiation source itself is not the source of artifact. Possibly the most common source of nonunif0rmities in the flood image is malfunction of one or more of the photomultiplier tubes in the detector head. Failure of a single tube will adversely affect not only events occurring in the field of view of the tube, but also events in the surrounding region because the loss of signal

from the defective tube results in miscalculation of the energy of an event adjacent to the defective tube. Such events are attributed as having a lower energy and are excluded from the energy window. Figure 5 shows a series of flood images from a system in which one edge tube had failed. The four images were acquired at different energies (energy window peaked at 140, 120, 100, and 80 keV). The defective tube covers only a small part of the useful field of view but causes the energy of nearby events to be underestimated. Figure 6 shows the effects of photomultiplier tube malfunction on a bone scan of the pelvis. A number of photon-deficient areas were seen in the right pelvis that did not match any known physical abnormality in the patient. A flood image acquired immediately after the scan confirmed that several photomultiplier tubes were out of tune. This example emphasizes that a -/-camera may malfunction at

Fig 4. Cold ring artifact (arrow) on intrinsic flood image caused by attenuation by the needle of the syringe containing the 99mTc.

Fig 5. Intrinsic flood image showing the impact on image uniformity, as a function of energy, of a defective photomultiplier tube at the edge of the field of view.

ARTIFACTS IN NUCLEAR MEDICINE

Fig 6. (A) Bone scan showing photopenic areas in the right pelvis. (B) Flood image acquired immediately after the bone scan showing cold regions corresponding to the photopenic regions on the bone scan, These were a result of malfunction of a number of photomultiplier tubes.

any time. Hence, acceptable image quality in the daily quality-control flood image does not ensure that the system will continue to function properly during the rest of the working day. Any unexpected findings in a clinical study should at minimum prompt a brief check of the photopeak setting and system uniformity to ensure that the finding is not the result of a system malfunction. A secondl less frequent cause of nonunifor-

259

mity in the flood image is a noisy photomultiplier tube. 6The location of a defective photomultiplier tube can usually be deduced from the pattern of nonuniformity in the flood image; however, a noisy tube often presents a confusing pattern. Figure 7A shows a vertical band of increased activity on the right side of the flood image, with a corresponding loss of resolution (Fig 7B). An image of point sources showed tails of activity on each of the sources (Fig 7C). The location of the noisy tube was diametrically opposite the nonuniform area and coincided with the intersection of imaginary lines drawn through the tails of the point sources. The tails were the result of random noise added to the pulse arithmetic signals by the defective tube. causing a shift in the positioning of some events toward the noisy tube. A more serious cause of nonuniformities in the ~/-camera flood image is a crack in the sodium iodide crystal. Sodium iodide crystals are extremely fragile and are easily fractured by mechanical or thermal stress. A sharp blow or a rapid temperature change (more than 3 ~ to 4~ per hour) can damage the crystal. Figure 8 shows the resulting fractures to the crystal of a dual-head whole-body system when some small 57Co plastic marker sources were inadvertently crushed between the collimator and the upper detector. Extensive damage of this nature requires replacement of the crystal. The light aluminum cover plate over the crystal generally offers little protection against bumps and scrapes, as can be seen in Figure 9, which shows the appearance of a small fracture caused by bumping part of the collimator against the cover plate. Damage to the crystal can be readily distinguishe d from other causes of nonuniformities by the presence of a cold defect with excess counts at the edge of the defect. These excess counts are the result of reflection at the fracture boundary ("edge packing" effect) and are characteristic of a fracture in the crystal. Apart from damage by mechanical or thermal stress, sodium iodid e crystals may also be damaged by hydratio n . Sodium iodide is highly hygroscopic, and the crystal is therefore hermetically sealed between the aluminum cover plate and the light pipe. AS a ~:camera ages, it is not uncommon for the hermetic seal to fail, allowing moisture to penetrate through to the crystal.

260

MICHAEL K. O'CONNOR

Fig 7, A no!sy photomult!plier tube on the left sideresulted in (A) a nonuniform band on the right side of the image, with (B) a slight 10ss of re~oluti~n. !C) Extrinsi~ image of pointsources showing "tails" On the sources; the "tails" increase in length from leftto right. (Reprinted by permission of the Society of Nuclear Medicine from: Oswald WM, O'Connor MK, A noisy photomultiplier tube: Its unusual effect on gamma Camera image uniformity. Journal of NuclearMedicine 1987; 15:157; figure 1.)

This results in discoloration Of the crystal and is generally evident in the flood images as a number of small discrete hot or cold areas. 7,8 Figure 10 shows an example of crystal hydration in a large-field-of-view system. The hydration has affected the outer rim of the field of view, and the central portion is unaffected. As the degree of hydration increases, more severe nonuniformities appear in the flood images, requiring eventual replacement of the crystal. In addition to problems with the crystal and photomultiplier tubes, a large number of electronic and mechanical problems can affect image quality. Figure 11 shows the flood image obtained after one of the circuit boards in the detector head became partly dislodged from its slot. Although the exact nature of such artifacts may not be easily attributable to a given component, the severity o f artifact they create usually is sufficient to prevent clinical use of the system and as such is not likely to corrupt the interpretation of clinical studies.

is apparent as nonuniformities in the flood image. Since the early 1980s most ~/-camera systems have incorporated online correction of this distortion (linearity correction) to ensure accurate positioning of the image data and to improve image uniformity. Before that time, many systems showed considerable distortion of line sources. These distortions, although artifactual. represented the limits of the technology at the time. On modern -/-camera systems, loss or corruption of the linearity correction map usually is evident as a significant deterioration in image uniformity and spatial linearity (Fig 12), although system resolution is largely unaffected. Updating the correction map usually rectifies this problem. Occasionally a failure in the

Resolution / Linearity The ability of a -y-camera to represent an object correctly i s degraded by a number of errors inherent i n the electronic and optical design of a ~-eamera The principle error affecting image resolution is Spatial distortion, which

Fig 8. Extrinsic flood images from the {A) lower head and (B) upper head of a dual-head whole-body syste m, showing extensive fractures to the crystal of the upper detector,

Fig 9. Small fracture to a sodium iodide crystal characterized by a cold defect with hot regions at the periphery of the defect,

ARTIFACTS IN NUCLEAR MEDICINE

Fig 10, Flood image showing early hydration of a sodium iodide crystal, characterized by a large number of hot spots at the periphery of the field of view.

analog-to-digital converter (ADC) may result in degradation or loss of resolution in one direction only. Figure 13 shows an example of a failure in an A D C resulting in poor resolution in the x direction while y resolution was unaffected. This type of problem may not be detected during the daily quality control because the flood image may not show a significant Change in uniformity, and the problem will be

Fig 11.

261

Fig 12. Image of an orthogonal hole phantom acquired

without linearity correction.

evident only if some type of resolution test pattern is imaged. Although system resolution and linearity can be assessed either extrinsically or intrinsically, extrinsic measurement may be confounded by interplay between lead bars in the resolution test pattern and the hole pattern in the coIlimatorz 9,1~ This is most evident with medium- or high-energy collimators, in which the septal thickness may approach the thickness of the bars in the test pattern. Figure 14 shows an example of a Moire interference pattern between a four-quadrant bar phantom and a m e d i u m - e n e r ~ collimator. Such interference patterns can be prevented by ensuring that all extrinsic measurements are performed using a low-energy collimator, in which the thickness of the septa is very much less than that of the bars in the test pattern:

Flood image showing loss of signal from a large

Fig 13. Images of a parallel line equal spacing (PLES)

number of photomultiplier tubes. This was caused by a loose circuit board in the detector head,

phantom showing poor resolution in the (A)x-direction Caused by partial failure of an analog to digital converter.

262

Fig 14. Corrupted image of a four-quadran t bar phantom. The Moire interference patterns are Caused by interplay between the collimator hole pattern and the bar spacing of the phantom.

Multiple- Window Spatial Registration Multipeak imaging with isotopes such as gallium 67 (67Ga) and indium 111 ( r a i n ) assumes that images acquired at different energies are directly superimposable. In conventional -y-cameras, the magnitude of the positioning signals generated by the pulse arithmetic circuitry increase with ~-ray energy.. These signals are normalized to some reference energy to ensure that the position of an event and. consequently, image size is independent of "y-ray energy. The accuracy to which this correction is achieved is termed the multiple-window spatial registration (MWSR) and is usually of the order of 1 to 3 ram. Larger values for the MWSR will result in poor coregistration of images acquired at different energies and lead to degradation in image resolution. Figure 15 shows the superposition of two images of a bar phantom acquired intrinsically using the 93-keV and 296keV photopeaks of 67Ga. Poor MWSR can be seen, both as a 10ss in resolution in the central part of the field of view and as a difference in the size of the 93-keV and 296-keV images at the edges of the field of view. In modern -c-cameras, the use of energy and linearity correction maps has significantly improved the MWSR. However, as a consequence, errors in the correction map or inappropriate use of a correction map for a given energy may lead to more subtle errors in clinical studies. Figure

MICHAEL K. O'CONNOR

Fig 15. Image of a PLES phantom acquired intrinsically using the 93-keV and 296-keV photopeaks of STGa. Differences in image size as a function of energy have resulted in loss of resolution in parts of the field of view.

16 shows the variations that may occur in the degree of misregistration between the 93-keV and 296-keV photopeaks of 67Ga over the field of view of a v-camera that was considered to be correctly

Fig 16. Subtraction image (296-kev image minus 93-keV image) showing an example Of the regional differences in the location of 296 keV (black) and 93 kev (white) point sources. (Reprinted by permission Of the Society of Nuclear Medicine from: Kelly BJ, O'Connor MK. Multiple Window spatial registration: Failure of the NEMA standard to adequately quantitate image misregistration with gallium-67. Journal of Nuclear Medicine. 1990;31:92-95; figure 2.)

ARTIFACTS IN NUCLEAR MEDICINE

263

ttmed for multipeak imaging. These variations cannot easily be characterized as a function of distance from the center of the y-camera and will lead to localized degradation in image quality,n If a large number of studies are being performed with isotopes such as rain and 67Ga, the MWSR should be checked on a monthly basis using the method described by the National Electrical Manufacturers Association (NEMA). 12 Collimators

Although y-camera collimators are often extremely heavy, their robustness is not in proportion to their weight, and the modern lowenergy, high-resolution collimator is very fragile and easily damaged. A small dent in the collimator will result in an apparent cold spot in the uniformity image (Fig 17). This type of artifact should not be confused with that caused by damage to the crystal (Fig 9) and can be distinguished by the absence of a bright rim around or adjacent to the cold spot. More extensive damage to a collimator, such as that shown in Figure 18, usually can be recognized by correlation of the defects seen on the flood images with the visible damage to the surface of the collimator. Although routine inspection of the collimator surface was possible with many older types of y-camera systems, newer systems often make it difficult to inspect the collimator because of patient crush protection pads attached to the front of the collimator. In such cases, damage

Fig 18. (A) Multiple scores on the underside of a collimator caused by failure to clear the locating pins on the detector head during collimator changes. {B) Resulting effects on the flood image.

Fig 17. Cold defect in a flood image caused by a small dent in the collimator,

can be deduced by comparing flood images obtained with different collimators. Collimators are manufactured by one of two methods, either in a mold (cast collimation) or by gluing together corrugated strips of lead (foil method). With the latter technique, mechanical stress or inadequate bonding of the adhesive may lead to a slight separation of the lead foils. The result is a crack along the length of the collimator that will be evident as a line of

264

increased activity on the flood images (Fig 19). Depending on the degree of radiation leakage through this crack and its location in the field of view, replacement of the collimator may be required. Other potential artifacts include poor mating of the collimator to the detector head and gaps between the collimator frame and the collimator itself. The flood image in Figure 20 shows hot spots at the edge of the field of view resulting from poor mating of the collimator to the detector head. Most collimator-related problems can be detected in the extrinsic uniformity images, a point in favor of extrinsic rather than intrinsic measurement of uniformity.

MICHAEL K. O'CONNOR

z

:i

i Z~7

Field of View In modern y-cameras, undesirable edge packing effects at the edges of the crystal usually are eliminated by electronically masking out the outer 1 to 2 cm of the useable field of view of the -/-camera. There are two potential problems associated with this electronic mask. The most insidious is incorrect adjustment of the mask radius, resulting in an unnecessary reduction in the active field of view of the -/-camera. We have seen instances in which a 50-cm field of view system:had an effective field of view of only 40 cm because of incorrect setting of this mask. Measurement of the distance between two point

Fig 20, Extrinsic flood image showing increased activity at the edges of the field of view caused by poor mating of the collimator to the detector head.

sources placed at the edge of the active field of view can be used to check the correct setting of the mask. A second, less common problem is misalignment of the mask with the actual field of view, resulting in visualization of the edge effects and reduction of the useful field of view (Fig 21). Verification of the field of view with a point source and a ruler is a simple check to perform, and one that should be done after any major service work on the system.

Whole-Body Imaging Whole-b0dy imaging using the older generation of ~-camera (40-cm field of view), required two passes along the length of the body. These two passes were then aligned to provide a complete whole-body image. Incorrect alignment of these two passes resulted in the well-

Fig 19. Crack in a foil collimator resulting in a line of increased activity running diagonally across the field of view.

Fig 21. Flood image showing misalignment of the electronic mask with the ~-camera field of view.

ARTIFACTS IN NUCLEAR MEDICINE

265

known vertical "zipper" artifact. The field of view of modern whole-body imaging systems generally is large enough (> 50 cm) to permit acquisition of a whole-body image with a single pass down the body. Although this eliminates the potential for the zipper artifact, it still requires movement of either the gantry or the imaging table, with the acquisition of a series of static views, or a continuous acquisition over the length of the body. With a series of static images, there is always the potential for incorrect alignment of the images resulting in a similar type of zipper artifact orientated horizontally across the whole-body image, such as shown in Figure 22. With continuous-mode acquisition, electronic timing errors or erratic motion of the gantry or patient table may result in banding artifacts in the whole-body image. Figure 23 shows an example of upper and lower whole-body flood images from a dual-head whole-body system. Electronic timing errors in the upper detector resulted in periodic loss of image counts and created the count-deficient bands seen in the upper part of Figure 23. In both of the above examples, the conventional flood images failed to indicate any problem with the system. These examples emphasize the need for occasional (every 1 to 2 months) checks of whole-body uniformity. This is a simple and rapid check to perform and can be done by placing a sheet source on the lower detector and acquiring whole-body flood images. 12

Hard Copy The traditional device for production of image hard copy has been the analog formatter. This is a essentially a cathode ray tube (CRT) combined with a series of lenses. When a valid event is detected by the -y-camera, an electronic timing signal turns on the CRT and illuminates the film with a single dot in a location corresponding to its detected position in the crystal. 13 Image artifacts can be introduced by a number of problems with this system. These include defocusing of the CRT dot, variations in dot intensity over the CRT face, and electronic timing problems resulting in streaking or incorrect positioning of the dot. If the CRT dot becomes defocused, the consequences are a blurring of the images and a decrease in apparent image intensity (Fig 24). This is often an

Fig 22. Whole-body image acquired as a series of static views, Minor mispositioning of the static images has resulted in horizontal lines of increased activity,

insidious type of problem because it may occur over a long period, and hence the loss in image quality may not be apparent from one day to the next. As CRTs age, there is often a variation in dot intensity over the CRT face. This translates into a variation in image intensity as a function of image position on film, which can result in underexposure or overexposure of an image.

266

MICHAEL K. O'CONNOR

Fig 23. Whole-body flood images from the (B) lower head and (A) upper head of a dual-head whole-body system. The bands of reduced activity on the flood image from the upper head were attributable to electronic timing errors causing a periodic loss in image counts.

Electronic timing problems often result in the superimposition of dots of varying size, shape, and intensity on film, resulting in the type of image seen in Figure 25. Becaus e most -y-cameras are now equipped with computer systems, problems with the analog formatter can be discerned easily by comparing the analog formatter image with that acquired on computer. With the demise of the analog formatter, most modern ~/-eamera systems rely on video formatters or digital printers for image hard copy. Artifacts related to these devices are easy to recognize because they do not appear on the soft copy (monitor) display of the images. These artifacts range from the monitor in the video

Fig 25. Flood image acquired on an analog formatter. Electronic timing problems have resulted in overexposure and incorrect placement of some of the dots,

formatter being out of focus or incorrectly synchronized to the video signal, to ghosting artifacts caused by incorrect impedance matching between the video signal from the computer and the video monitor. Digital printers or formatters tend to be more stable than their analog or video counterparts and generally are an all-or-nothing affair in terms of image production. A secondary cause of artifacts in the hard copy process is the film processor itself. This

Fig 24. Images of a PLES phantom acquired (A) on an analog formatter that was out of focus and (B) after refocusing of the formatter.

ARTIFACTS IN NUCLEAR MEDICINE

may introduce a wide variety of streaks, blemishes, and other disfiguring marks on the final hard copy. Figure 26 shows an example of streaks on a flood image that could be misinterpreted as a problem with the `/-camera or collimator. Fortunately, processor-related artifacts usually are easy tO recognize because they often extend beyond the edges of the images, are independent of the -/-camera used for image acquisition, and occur in a relatively reproducible location on film each time.

Computer With the integration of computers into all aspects of -/-camera technology, the potential exists for the generation of a large variety of artifacts resulting from the failure of digital components of the system. Problems can range from corruption of the energy, linearity, and uniformity correction maps to failure of some stage of the digitization process to failure to correctly display tile digital image. The digital nature of these artifacts often allows then to be recognized as such. Figure 27 shows an example of a corruPted uniformity Correction map and its effect on image quality. The digital nature of the artifact Can be recognized easily from the appearance of discrete pixels in the image. Figure 28 shows another example of count losses resulting from the failure of a memory chip in a uniformity correction module. The

267

increasing amount of image processing in modern -/-camera systems can result in a large number of digitally related artifacts. Although it is not feasible to review all the causes of such artifacts, the digital nature of these artifacts often is a clue to their origin. In the current generation of -/-camera systems, most computerrelated artifacts show properties that allow the user to distinguish them from those related to the patient, radiopharmaceutical, or detector head. As the latest digital systems permeate into the clinical workplace, we are likely to see a new generation of image artifacts that may require modifications or additions to how we currently Check or perform the quality control on `/-camera systems. TOMOGRAPHIC IMAGING

Introduction It is well recognized that tomographlc systems are more complex than planar systems, and the disparity in complexity between the two modes of operation has increased in recent years with the advent of multidetector systems and is likely to increase further as new advances such as scatter correction, attenuation correction, and coincidence imaging move from the research environment into routine clinical use. Full evaluation of the possible sources of artifacts in a modern tomographic `/-camera and computer system may involve performing a large number of sophisticated tests of system function, many of which require specialized test equipment and are beyond the capabilities of most small laboratories. This section reviews the most common artifacts and describes some simple checks that can aid in their detection.

Uniformity

Fig 26. Lines of increased intensity across the flood image caused by dirty rollers in the film processor,

One of the major sources of artifacts in tomographic imaging is a nonuniformity in the flood image. The degree of nonuniformity in the response of a `/-camera to a uniform field of radiation can be defined in terms of the global and local variations in uniformity over the field of view (integral and differential uniformity, respectively). 12 Integral uniformity is defined as the largest variation (maximum minus minimum) in counts over the useful field of view, and differential uniformity is a measurement of the worst-case rate of change of uniformity over

268

MICHAEL K. O'CONNOR

Fig 27. Flood images from a small-field-of-view ~-cam~ra system (A) before and (B) after application of uniformity correction.

a limited distance (5 pixels). Modern ~-camera systems typically have integral and differential uniformities of between 4% and 7%. Nonuniformities of this magnitude can generate ring artifacts in tom0graphic data. 14 hence all tomographic systems apply an additional correction to the raw image data, called "uniformity," "flood," or "sensitivity" correction, before data reconstruction. After uniformity correction, a tomographic system in good working order will have values of differential uniformity in the range of 1.0% t o 2.5%, with values of integral uniformity a little higher at 1.5% to 3.5%. To understand t h e magnitude and type of nonuniformity that can create a ring artifact, it is necessary to understand the mechanism by which ring artifacts are generated. The recon-

struction process assumes that the ~-camera has perfect uniformity. During back-projection, all pixels are assumed to have equal weight and are back-projected with equal intensity. If the counts in one pixel of the image are artificially reduced (eg, because of a dent in the collimator), information at that location will be back-projected at that reduced level. The result in the reconstructed image will be a ring artifact, with the radius of the ring equal to the distance of that pixel from the center of rotation. 15 In practice, no system shows perfect uniformity, and there will always be minor changes in uniformity from one pixel to the next, even with uniformity correction. Each and every one of these minor changes in uniformity will result in a ring artifact. However, they are of concern only when their magnitude exceeds the noise level in the transaxial image and they can be perceived above image noise. Figure 29 shows two 30 million count flood images that were used to create uniformity correction maps. Both images were acquired using a refillable sheet source containing 99mrc. In one case, the Sheet source

P|g 28. Vertical bands of reduced counts in a bone scan. This artifact was caused b y failure of a memory chip in the uniformitY correction module of the ~-camera system.

Fig 29, Thirty-million count flood images acquired using (A) a flat 99mTc sheet source and (B) a bulging 99mTc sheet source created by overfilling of the sheet source.

ARTIFACTS IN NUCLEAR MEDICINE

Fig 30. Corresponding transaxial slices of a uniform elliptical phantom reconstructed using correction maps generate d from the flat and bulging flood !mages shown in Fig 29.

was kept perfectly flat and yielded a very uniform flood image, with values of 1.9% and 1.2% for integral and differential uniformity, respectively. In the second case, the source was overfilled, resulting in a bulging source with a hot center and cold edges, giving values of 12.1% and 1.8% for integral and differential uniformity, respectively. Figure 30 shows transaxial images of a Jaszczak phantom that were corrected for uniformity with the two images in Figure 29. Although there were significant nonuniformities in the bulging flood image, they did not translate into ring artifacts in the transaxial imag e . The reason for this is the small change in uniformity from one pixei to the next in the buig!ng image, ie, small values of differential uniformity. Therefore, the magnitude of the resulting ring artifacts is correspondingly small. If the nonuniformities in the refillable sheet source were concentrated over a small area, for example, resulting from an air bubble, the large change in uniformity over a few pixels (ie, high value of differential uniformity) would have resulted in significant ring artifacts. Therefore,

Fig 31. Transaxial slices reconstructed in an identical manner from two studies of a uniform elliptical phantom. Total counts were (A) 5 x 10e and (B) 100 x 10 6.

269

for tomographic imaging measurement of differential uniformity is the most useful indication of the suitability of the system for single photonemission computed tomography {SPECT). How good (o r low) the values of differential uniformity must be to ensure the absence of ring artifacts in clinical studies depends on the clinical study. The purpose of uniformity correction is to reduce system nonuniformities to a low enough level that any ring artifacts generated from residual nonuniformities in the system are less than image noise and hence are not visible in the reconstructed data. However, image noise in the reconstructed data will vary significantly with the type of study, hence a low-count 111In SPECT study will be far noiser than a high-count 99mTc liver SPECT study. Therefore, a ring artifact that may be clearly visible in the 99rnWcliver SPECT study may not be visible in the ~11InSPECT study~ This implies that greater nonuniformities can be tolerated for some types of clinical studies than for others: 16 Rather than estimating the level of system uniformity required for different types of studies, it is simpler to correct system nonuniformities to a level that will ensure no visible ring artifacts in even the highest-count clinical studies. In general, it has been found that a 30 million count flood image provides sufficiently good correction of system nonuniformities for all current clinical studies. This correction should result in a value for differential uniformity of less than 3%. ~6 This level of correction may not be adequate for some types of phantom studies. For ex-

270

MICHAEL K. O'CONNOR

Fig 32. Sagittal slice from a brain scan showing hot spots in the.superior cerebrai Cortex and basal ganglia (A) caused by ring artifact s. (B) The Corrected slice is shown. (Reprinted by permission of the Society of Nuclear Medicine from: O'Connor MK. Corruption of brain SPECT studies caused by error in uniformity correction algorithm. Journal of Nuclear Medicine 1990;31:700-701; figure 1.)

ample, a Jaszczak phantom with 10 mCi (370 MBq) Of 99mTcmay give 50 to 70 million counts Over a 30-minute S P E C T acquisition, and in such studies ring artifacts may be noted even though the system uniformity is adequate for Clinical studies. To demonstrate this point, Figure 31 shows 1-pixel-thick transaxial Slices from tWO p h a n t o m studies acquired and reconstructed in an identical manner. The only differ-

no artifact

ence was an increase in the time per frame by a factor of 20, giving total counts of 5 million and 100 million in the two studies. The low-count study was noisy but free of ring artifacts, and the high-count study showed ring artifacts in all slices. Although the figures above are textbook examples of ring artifacts, in clinical practice such artifacts may be difficult to recognize, particularly if they occur on or near the center Of rotation. Figure 32 shows two sagittal images Of the brain. The top image contains two hot spots that were ring artifacts caused by a software error in the application of the uniformity correction map. 17 The lower one shows the same image after application of the proper correction map. In cardiac imaging, the use of oblique slices makes it difficult to recognize a ring artifact because it no longer resembles a ring in shape. Figure 33 shows how misleading a ring artifact can be i n a clinical study. This artifact manifested itself as a decrease in the uptake of 99mTc-sestamibi in the septal wall of the heart. Although identification of such an artifact can be difficult, a review of the transaxial slice images can often help, Figure 34 shows one of the transaxial images from the study shown in Figure 33, with the ring artifact now visible as a partial ring (arrow). Placement of a cross-hair in the center of the image shows the ring to be concentric with the cross-hair, indicating that it is likely to be a ring artifact. It is helpful to remember that ring artifacts are always concentric with the center of rotation, and hence Should be concentric with the center of the transaxial slice (unless the data

ring artifact

Fig 33. Short-axis slices of the heart before and after correction for a ring artifact, The artifact caused an apparent reduction in septal wall activity,

ARTIFACTS IN NUCLEAR MEDICINE

271

Fig 34. Transaxialimage from the study shown in Fig 33. The ring artifact (arrow) is now easily identified and is concentric with the center of the image (cross-hair).

have been reconstructed with a zoom and offset).

Rotational Uniformity On most SPECT systems, the "uniformity" correction map is generated from a flood image acquired with the detector head facing up. It is then assumed that this correction map can be applied to images acquired at all angles of rotation. There are a number of instances in which this assumption may not be valid. The photomultiplier tubes within the detector head are heat sensitive; therefore, heat generated by the electronics within the head may alter their characteristics. If the heat distribution within the head changes as a function of angle, then so too can the performance characteristics of the detector. Another likely cause of variations in uniformity with angle of rotation is poor optical coupling of a photomultiplier to the light guide or crystal. This can cause the tubes to decouple slightly at certain angles, leading to a significant change in uniformity. Photomultiplier tubes are also sensitive to the earth's magnetic field. Their performance characteristics can change with their orientation to the earth's magnetic field, leading to changes in uniformity with rotation. 14 This may be a problem in older SPECT systems with inadequate magnetic shielding. A simple test for the assessment of rotational uniformity is to securely tape a lightweight 57Co sheet source to the collimator face and perform a high-count tomographic acquisition over 360~ The raw'data can then be played back in cine mode to' check for significant changes in uniformity with rotation. Figure 35 shows an example of the effects of heat changes in the detector

head on image uniformity. Such variations are difficult to correct and if severe enough ( > 2% change in uniformity with rotation) may require a redesign of the detector head. A more common finding is decoupling of a photomultiplier tube, as shown in Figure 36. This problem occurred in a 6-year-old system after the optical coupling grease had dried out. On older systems (> 10 years), magnetic shielding is often inadequate. Figure 37 shows an example of variations in image uniformity caused by poor shielding. This problem is rarely seen on modern systems unless a Mu metal shield is inadvertently omitted. Measurement of rotational uniformity should be performed once or twice a year, or whenever a significant upgrade or repair has been made to the detector head.

Center of Rotation Accurate center of rotation (COR) correction is important for high-quality tomography. Errors in COR of as little as 0.5 pixel in a 128 x 128 matrix can lead to degradation in image

Fig 35. High-count flood images acquired every 45 ~ over 360 ~ Heat changes in the detector head resulted in changes in image uniformity as a function of angle.

MICHAEL K. O'CONNOR

272

0-36

180-216

36-72

72-108

216-252

252-288

108-144

288-324

144-180

324-360

Fig 36. Rotational uniformity--lO images acquired over 360 ~ showing a change in system uniformity between 180 ~ and 252 ~ because of decoupling of one of the photomultiplier tubes.

quality. 18 Figure 38 shows the effects of a 1 pixel (3.2 mm) error in C O R on transaxial images of a Jaszczak phantom containing hot rods. The error can be recognized easily by the loss of resolution and the misalignment of the 6 pie segments of the phantom. Unfortunately, as with uniformity, errors in COR in clinical studies may be more difficult to recognize. Figure 39 shows examples of horizontal long axis slices of the heart reconstructed with different COR values. An error in C O R often results in the appearance of an apical defect in the myocardium. To determine whether such a defect is the result of a COR error, it is again helpful to review the transaxial images. If a C O R error is present, these will show tails of activity coming

Integral Uniformity = 3.0 % @ 0 degrees

off the apex of the heart, as seen in Figure 40 (arrows). Not only is it important to use the correct value of COR, it is also essential that this value remain constant as a function of angle. When measured on a ~-camera system, at a radius of rotation of 20 cm, both the x and y values for the COR should show less than a 2-ram variation over a 360 ~ orbit. C O R normally is a very stable parameter of modern -y-camera systems, and a weekly check is adequate to ensure proper correction.

Collimator Hole Alignment Parallel-hole collimators are assumed to have the holes oriented perpendicular to the surface of the crystal. In reality, there can be consider-

Integral Uniformity = 5.1% @ 90 degrees

Fig 37. Change in system uniformity caused by inadequate magnetic shielding around the photomultiplier tubes. {A) Integral uniformity = 3.0% @ 0~ (B) Integral uniformity = 5.1% @ 90~

ARTIFACTS IN NUCLEAR MEDICINE

273

shows the range of variations in C O R for approximately 170 collimators tested in our laboratory. Approximately 16% of collimators were considered unsuitable for tomographic imaging. Measurement of collimator hole alignment only needs to be performed once for each collimator because it is solely a function of the physical characteristics of the collimator. Fig 38. Transaxial slice through the hot rod section of a Jaszczak phantom reconstruction with [A) the correct COR value and with (B) a 3.2-ram error in COR.

able variation in collimator hole angle, both locally and globally. These variations directly affect the COR, and a poorly manufactured collimator may have considerable variations in COR across its surface. These variatioris cannot be corrected for by the standard C O R correction method, and their presence cannot be determined by inspection of planar image quality. While such collimators often produce acceptable planar images, the quality of SPECT images will be inferior. Unfortunately, the only solution for this type of problem is replacement of the collimator. A simple way to check a collimator is to perform multiple measurements of C O R along the length of the collimator, and check that all measurements essentially give the same result, a8 Maximum variations in C O R should not exceed _+1.0 mm over the length of the collimator.! 9,2~ Larger variations than this indicate local hole angulation errors. Furthermore, the average COR value for the collimator should be similar to those for the other collimat o r s on the tomographic system. Figure 41

- 1.5 pixels Fig 39.

GantryAlignment (Single-HeadSystem) For a tomographic acquisition in a singleheaded SPECT system, the gantry usually is set to 0 ~ and the detector head is leveled before an acquisition. Setting the detector head level is based on the assumption that the axis of rotation of the detector head is horizontak This axis of rotation is determined by the alignment of the gantry. Misalignment of the gantry can be caused by a number of things. In many older SPECT systems, sagging of the detector arms can occur. This is particularly true in cantilevered systems as compared with counterbalanced systems. The gantry itself may not be level on the floor because of either incorrect shimming of the gantry or irregularities in the surface of the floor. Whatever the cause, the consequences are that leveling the detector head results in data being acquired obliquely rather than perpendicular to the axis of rotation. For example, a 1~ gantry misalignment with the detector head at a 20-cm radius of rotation will cause a 3.5-ram displacement of image data along the axis of rotation. Variations in gantry alignment will be visible in the analysis of the weekly COR measurement as variations

0 pixels

1.5 pixels

Effect of errors in the center of rotation on horizontal long-axis slices through the heart (1 pixel = 6.4 mm).

274

MICHAEL K. O'CONNOR

- 1,5 pixels

0 pixels

1.5 pixels

Fig 40. Transaxial images from the study shown in Fig 39. In cardiac studies, a COR artifact can be identified by the characteristic tail (arrows) that emanates from the apex of the heart.

in the y-position of the point source. Figure 42 shows the distortion introduced into a transaxial image of the heart by a 5 ~ gantry misalignment error. Gantry alignment and its stability with rotation can also be checked easily using a small bubble level. The gantry should be leveled at 0 ~ then rotated through 180 ~ and checked to ensure that the gantry is still level. Alignment should be checked once or twice a year and after any major upgrade or modification to the gantry.

Head Alignment (Multihead Systems) Many multidetector systems do not permit tilting of the detector head. and hence if the gantry is not vertical it is not possible to level the head and misalign the detector relative to the gantry. However, a new concern with multidetec120 100

Accept i Reject

tor systems is the alignment of a given detector to itself at different radial positions and the alignment between the different detectors. Radial alignment can be checked by imaging a line source at two different positions. Figure 43 shows images of a line source acquired at distances of 5 cm and 30 cm. Subtraction of the 5-cm image from the 30-cm image indicates whether the two images overlap precisely (Fig 43C). Any misalignment of the detector with radial motion will be easily apparent on the subtracted image (Fig 43D). This test should be performed in both the x and y directions. This is a very simple but useful test that will detect not only problems with radial alignment, but also any errors in collimator hole alignment. For interhead alignment, it should be noted that most systems contain corrections for head alignment (these should have been created and loaded by the service engineer). These corrections should be up to date and active. The

o

o

o

80

N=168

60

Accept = t41 (84 %) Reject = 27 (16 %)

40 Z

20 0

<1

1-2 2-3 3-4 4-5 Maximum change in COR (ram)

>5

Fig 41. Maximum change in COR over the face of the collimator. Results from the evaluation of a series of 168 collimators, Maximum acceptable change is 2 mm at a 20-cm

radius of rotation,

Fig 42. Transaxial images of a cardiac phantom with (A) correct alignment and with (B) a 5~ alignment error. Note the increase in apical activity and loss of counts in the inferior wall of the heart (arrow).

ARTIFACTS IN NUCLEAR MEDICINE

275

Fig 43. Images of a line source acquired by radially moving the detector from a distance of (A) 30 cm to (B) 5 cm from the collimator face. Good radial alignment is seen in C, whereas a 2-ram misalignment is present in D.

accuracy 'of the corrections can be checked by acquiring tomographic images of a test pattern a n d reconstructing data from each head separately. These results should then be compared with those obtained by reconstructing data from all the heads. Once correctly installed, the head alignment correction should not change significantly over time because it is a function of the mechanical movement of the gantry and need only be checked on a yearly basis. Fan-Beam Collimators Fan-beam collimators are coming into widespread use for brain and cardiac studies. The cause of artifacts associated with this type of collimator is often difficult to determine because the raw data are in a distorted form and difficult to interpret. One potential problem associated with fan-beam collimators is failure of the reconstruction software to recognize the data as fan-beam. This leads to highly distorted reconstructed data. Similar but less severe artifacts can occur if incorrect values are used for the fan-beam focal length. Figure 44 illustrates

the effects of the former problem on the quality of transaxial slices through a Jaszczak phantom. A simple way to evaluate a fan-beam collimator is to reconstruct images of a point source. The effects of changes in the fan-beam parameters can then be seen as changes in the resolution of the point source. These parameters need only be determined once because they are a function of the geometry of the collimator. Miscellaneous Items A number of other factors can potentially cause artifacts in tomographic studies. These include other system components, such as the tomographic table, as well as environmental factors such as temperature and adjacency to other sources of radiation or electromagnetic fields. Tomographic tables are designed to have low attenuation characteristics. However, many tables contain embedded metal brackets, screws, or supporting rods (Fig 45). The location of these items should be ascertained by a simple transmission scan through the t a b l e . If such

Fig 44. Transaxial images of the cold rod section of a Jaszczak phantom. Image data were acquired using fan-beam colnmator and reconstructed with (A) the correct focal length and with (B) the focal length set to infinity (parallel hole collimation).

276

MICHAEL K. O'CONNOR

10

o~ E

o ."~8

6

19

20

21 22 23 24 25 26 27 Room Temperature (Degrees C)

28

29

Fig 47. Effect of changes in room temperature on the measured value of system uniformity. The system was calibrated at 20~

Fig 45. Transmission image through the end of a tomographic table. The supporting bracket for the head rest shows high attenuation, and patients should not be imaged over this portion of the table.

items are present, the appropriate region of the table should be marked and patients positioned such that this region is outside the area being imaged. Although all tomographic tables show low attenuation when imaged face on, they may show significantly higher attenuation at other angles. Figure 46 shows the variation in counts 110 100

v-r . . . .

........... >.,

[ ~ ' | ~

ill

.....

1

E

"II,

50

Ii !/

Nl /~ iI

V 40 60

120 180 240 Angle of Rotation, degrees

300

360

Fig 46. Variation in counts transmitted through a variety of tomographic tables as a function of angle of rotation of the detector for a 2~ line source placed at a height of I cm above the center of the table. (Reprinted by permission of the Society of Nuclear Medicine from O'Connor MK, Bothum ED: Effects of tomographic table attenuation on prone and supine cardiac imaging. Journal of Nuclear Medicine 1995; 36: 1102-1106; Figure 3).

transmitted through different tomographic tables as a function of angle. The degree of attenuation is highest when the table is viewed at an angle of 90~ This attenuation can artifactually alter the apparent distribution of counts in a tomographic study. Although most tomographic tables do not significantly affect image quality, those constructed of aluminum or showing a large radius of curvature (ie, very flat tables), can alter the apparent count distribution in the heart during tomographic imaging, particularly if patients are imaged in the prone position. 21 All -/-camera manufacturers require that room temperature be kept within certain limits (typically 10~ to 25~ and that temPerature changes be less than 3~ per hour to protect the integrity of the sodium iodide crystal. However, even within these limits, changes in room temperature over the working day may be sufficient to alter system performance. Photomultiplier tubes are temperature sensitive, and changes in room temperature may alter theirgain, resulting in a change in the location of the photopeak. Figure 47 shows the effect of changes in room temperature on the integral uniformity of a large-fieldof-view system. Although many modern systems incorporate some type of autotune feature to compensate for such effects, older systems-may lack such features, and a stable operating environment is essential in such cases. CONCLUSIONS

As the complexity of modern systems increases, it becomes more important that both

ARTIFACTS IN NUCLEAR MEDICINE

277

the technologist and the physician be able to recognize the various types of artifacts that can occur in ~-camera systems and their potential impact on clinical studies. A thorough evaluation of the system at installation and a comprehensive quality-control program will detect the majority of problems that can occur. With the improved stability of modern systems and the latest generation of "smart" detector systems that have streamlined and automated much of

the image acquisition and analysis, it is easy for the user to become complacent and assume that the system will "take care of everything." The failure of a system component can occur at any time, and in modern systems may be subtle and difficult to recognize. In such an environment it is imperative that any unexpected finding in a clinical study be questioned with respect to a possible malfunction of some aspect of the acquisition and analysis of the data.

REFERENCES 1. Chandra R: Operational characteristics and quality control of imaging devices, in Introductory Physics of Nuclear Medicine (ed 3). Philadelphia, PA, Lea and Febiger, 1987, pp 165-180 2. Graham LS: Quality assurance of anger cameras, in Rao DV, Chandra R, Graham MC (eds): Physics of Nuclear Medicine--Recent Advances. New York, NY, American Institute of Physics, 1984, pp 68-82 3. Cranage RW, Peake JCF: The effect of high energy impurities on measurements of gamma-camera resolution and uniformity using Co-57 flood sources. Br J Radiol 52:81-82, 1979 4. Busemann Sokole E, Kugi A, Bergmann H: High count rates cause non-uniformities in cobalt-57 flood images. Eur J Nucl Med 20:896, 1993 5. English RJ, Polak JF, Holman BL: An iterative method for verifying systematic nonuniformities in refillable flood sources. J Nucl Med Technol 12:7-9, 1984 6. Oswald WM, O'Connor MK: A noisy photomultiplier tube: Its unusual effect on gamma camera image uniformity. J Nucl Med 15:157, 1987 7. Paras P, Van Tuinen R, Hamilton Di Quality control for scintillation cameras, in Rhodes B (ed): Quality Control in Nuclear Medicine. St Louis, MO, Mosby, 1977, pp 336-348 8. Lukes SJ, Grossman LW, Nishiyama H: Thallium-201 imaging artifacts not detected by technetium-99m or cobalt-57 quality coritrol testing. Radiology 146:237-239, 1983 9. Bonte FJ, Graham KD, Dowdey JE: Image aberrations produced by multichannel collimators for a scintillation camera. Radiology 98:329-334, 1971 10. YehEL: Distortionofbar-phantomimage bycollimator. J Nucl Med 20:260-261, 1979 il. Kelly BJ, O'Connor MK: Multiple window spatial registration: Failure of the NEMA standard to adequately

quantitate image misregistration with gallium-67. J Nucl Med Technol 18:92-95, 1990 12. NEMA: Performance measurements of scintillation cameras. Washington, DC, National Electrical Manufacturers Association (NEMA), 1994, Standards Publication No. NU 1-1994 13. Weber DA, Ivanovic M: Display devices, in Simmons GH (ed): The Scintillation Camera. New York, NY, Society of Nuclear Medicine, 1988. pp 60-78 14. Rogers WL. Clinthorne NH. Harkness BA, et ah Field flood requirements for emission computed tomographywith an Anger camera. J Nucl Med 23:162-168, 1982 15. Gullberg GT: An analytical approach to quantify uniformity artifacts for circular and noncircular detector motion in single photon emission computed tomographic imaging. Med Pbys 14:105-114, 1987 16. O'Connor MK, Vermeersch C: Critical examination of the uniformity requirements for single-photon emission computed tomography. Med Phys 18:190-197. 1991 17. O'Connor MK: Corruption of brain SPECT studies caused by error in uniformity correction algorithm. J Nucl Med 31:700-701, 1990 18. Cerquira MD. Matsuoka D. Ritchie JL, et al: The influence of collimators on SPECT center of rotation measurements: Artifact generation and acceptance testing. J Nucl Med 29:1393-1397. 1988 19. Busemann-Sokole E: Measurement of collimator hole angulation and camera head tilt for slant and parallel hole collimators used in SPECT. J Nucl Med 28:1592-1598. 1987 20. Malmin R. Stanley P, Guth WR: Collimator angulation error and its effect on SPECT. J Nucl Med 31:655-659. 1990 21. O'Connor MK, Bothun ED: Effects of tomographic table attenuation on prone and supine cardiac imaging. J Nucl Med 36:1102-1106, 1995