Routine Quality Control of Clinical Nuclear Medicine Instrumentation: A Brief Review

γ-Camera uniformity may be evaluated either intrinsically or extrinsically. Intrinsically, point source is placed at least 3, and preferably 5, crystal dimensions from and centered over uncollimated detector to provide uniform photon flux (left). Extrinsically, uniform flood, or sheet, source of ⁵⁷Co is placed directly on collimated detector (right).

Sources of γ-camera nonuniformity. (A) Mistuning (or detuning), meaning that photopeak of radionuclide does not coincide with photopeak energy window of camera, perhaps because energy window (as shown) or high voltages of PMTs are not set correctly. (B) Uncoupling of PMT from crystal, resulting in loss of all or part of light signal in resulting air gap between PMT entrance window and crystal. (Courtesy of Dr. Barbara Binkert, New York Presbyterian Hospital, New York, NY.) (C) Cracked crystal, either because of mechanical trauma (impact) or temperature excursion (i.e., temperature increase or decrease at rate faster than ∼5°C per hour, causing crystal to expand or contract, respectively, to point of cracking). Note that it is rate of temperature change that is critical. Photographs on right show cracked crystal that produced corresponding image. Even though cracks are grossly imperceptible, artifacts produced are dramatic. (Courtesy of Dr. Barbara Binkert, New York Presbyterian Hospital, New York, NY.) (D) Corrupted, deleted, or switched-off software correction tables. Even perfectly functioning γ-cameras have some nonuniformity due to point-to-point variations in energy spectra, greater sensitivity at and lower sensitivity between PMTs, and residual nonuniformity due to ill-defined factors such as variations in crystal thickness. Associated nonuniformities are measured and used to create energy, linearity, and uniformity (or sensitivity) correction tables. Note that linearity correction table has biggest effect on uniformity: if corrupted, deleted, or switched off, PMT pattern becomes grossly apparent, and IU approaches 20%. Fortunately, in contrast to uniformity correction table and, to lesser extent, energy correction table, linearity correction table rarely needs to be updated once γ-camera is installed; if updating becomes necessary, it is almost always done by field-service personnel of manufacturer, not by end-user.

(A) Photograph and γ-camera image of 4-quadrant bar phantom, with schematic diagram illustrating setup for assessment of spatial resolution using such phantom. (Though seldom used in practice, sheet source that may be filled with different radionuclides and used to evaluate extrinsic uniformity is also shown in photograph.) (B) Energy spectrum for 662-keV γ-ray emitted by ¹³⁷Cs, illustrating definition of energy resolution as percentage FWHM of photopeak energy, Eγ.

(A) COR misalignment and resulting image-blurring artifacts in rotating-γ-camera SPECT. Degree of blurring is related to magnitude of spatial misalignment of mechanical and electronic CORs. Misalignment as small as 3.2 mm (or 0.5 pixel for 64 × 64 image matrix) can produce perceptible blurring in SPECT images, with blurring substantially worse for misalignment of 6.4 mm (1 pixel). (Adapted from reference (29) with permission.) Note that for cross-sectional image of line source, COR misalignment blurs expected point into full or partial circle depending on position of source in FOV: if it is at or near center of FOV, line source appears as full circle in cross-section; if it is near periphery of FOV, it appears as partial circle. (B) COR misalignment can be measured and corrected on basis of acquiring 360° circular SPECT study of ^99mTc point source and constructing graphs of x- and y-positions (perpendicular and parallel to axis of rotation, respectively) of position of maximum-count pixel in each projection image vs. angular position. x- and y-position vs. angle graphs should be sinusoidal curve and straight line, respectively. Angle-by-angle deviation between x-position on best-fit sine curve and x-position of actual maximum-count pixel thus yields correction table, indicating offset by which each projection image must be shifted at each angular position to align CORs. Alternatively, average of offsets may used at each angular position. (Adapted from reference (15) with permission.)

(A) Severe ring, or bull's-eye, artifact in transverse SPECT image through 20-cm diameter ^99mTc-filled cylinder phantom resulting from gross γ-camera nonuniformity (i.e., IU ≈ 10%). (B) Appearance of such artifact (arrow) in clinical SPECT image, transverse image from ^99mTc-sulfur colloid liver–spleen study. (C) In contrast to A, this is acceptable transverse SPECT image through ^99mTc-filled cylinder phantom. No perceptible ring, or bull's-eye, artifact is demonstrated; IU at time of this acquisition, ∼2%, was well within tolerance. In cylinder phantom images, A and B, analytic postprocessing attenuation correction was applied.

Photographs of disassembled (end view) (A) and assembled (side view) (B) phantom (Deluxe Jaszczak Phantom; Data Spectrum Corp.), used for evaluation of overall performance of tomographic imaging systems. This fillable acrylic phantom is 22 cm in diameter by 19 cm in length and includes plain section for evaluation of tomographic uniformity (C), section containing empty (cold) spheres ranging from 9.5 to 31.8 mm in diameter for evaluation of cold-sphere contrast (D), and section containing solid (cold) rods ranging from 3.2 to 11.1 mm in diameter for evaluation of reconstructed spatial resolution (E). Images in C–E have been analytically corrected for attenuation. (Courtesy of Data Spectrum Corp.)

(A) Sinogram (i.e., histogram) presentation of emission tomography (i.e., SPECT or PET) data. (B–D) PET sinograms of uniform-cylinder source without any visually perceptible discontinuities or other artifacts (B), with blank diagonal line indicative of faulty detector (crystal) element (C), and blank diagonal band indicative of faulty detector block (D).

Two-dimensional PET sinograms and reconstructed transverse images through hot-sphere ¹⁸F-filled phantom: sinogram without any perceptible artifacts (A) and corresponding transverse image (B); sinogram with blank diagonal band, suggestive of faulty detector block (C) and corresponding transverse image (D). Despite obvious defect in sinogram C, reconstructed images B and D are virtually indistinguishable. (Courtesy of Dr. Osama Mawlawi, M.D. Anderson Cancer Center, Houston, TX.)

Proprietary graphical display (GE Healthcare), detector block–by–detector block, of relative values of PET scanner operational parameters derived from blank scan, including coincidence counting rate, singles counting rate, detector dead time, coincidence timing window, and energy setting. Such display allows operator to quickly and easily discern out-of-tolerance results, displayed with grossly different (i.e., lower) intensity than within-tolerance results. (A) Display for acceptable blank scan; that is, blank scan for which all detector parameters are within tolerance. (B) Display for blank scan in which coincidence counting rate, singles counting rate, and dead-time results for 1 block detector are out of tolerance, indicated by black areas (arrows) in respective displays. (Courtesy of Dr. Osama Mawlawi, M.D. Anderson Cancer Center, Houston, TX.)

Reconstructed coronal images of ⁶⁸Ge uniform-cylinder phantom without (A) and with (B) normalization applied. (A) Unnormalized (i.e., uncorrected) image has obvious artifacts attributable to differences in sensitivities between direct and cross planes and presence of separate rings of block detectors. (B) Appropriate normalization virtually eliminates these and other artifacts related to nonuniformity of scanner response.

Cross-sectional diagrams of GE quality assurance phantom (24) illustrating respective sections (inserts) for evaluation of laser-light alignment, image slice thickness, spatial resolution, linearity, and high-contrast contrast resolution (A); low-contrast uniformity (B); and image uniformity and noise (C). Various short black lines in acrylic insert in section shown in A are cavities that fill with water when phantom is filled, providing high contrast between cavities and acrylic; portion of this section outside this insert, although not shown in black, is also water-filled. Diagonally arranged sets of line cavities ranges are 1.6 (lower left), 1.3, 1.0, 0.8, 0.6, and 0.5 (upper right) mm in line width and are used to evaluate spatial resolution. Section shown in B includes polystyrene membrane with series of holes (10, 7.5, 5, 3, and 1 mm in diameter) that also fill with water when phantom is filled, providing low contrast between holes and polystyrene. (D) Side-view diagram (not to scale) of section of phantom in A, showing only 1 of slice-thickness measurement components of insert; these line cavities are air-, not water-, filled. They are staggered (offset) 1 mm apart in longitudinal direction. Also shown in D, as well as in A, is laser-alignment groove around circumference of phantom and 2 corresponding laser-alignment cavities. Boxes indicate pertinent ROIs for different analyses. (See text for details.)

Setup for measurement of CT radiation exposures using ionization chamber and 32-cm-diameter acrylic torso (body) phantom. (Courtesy of Dr. Sadek Nehmeh, Memorial Sloan-Kettering Cancer Center, New York, NY.)