Abstract
Objective:The purpose of this study was to determine experimentally the minimum thickness (Dmin) of a defect inserted on the myocardial wall of a cardiac phantom at different locations that could be clearly detected in a SPECT perfusion study using 99mTc and 201Tl.
Methods:Rectangular (or cylindrical) defects with the same thickness were inserted on the inner surface of a myocardial phantom at 5 different locations: anterior (ANT), septal (SEP), inferoposterior (IP), lateral (LAT), and apical (AP). For different defect thickness (from 1 to 7 mm, in increments of 1 mm) the myocardial SPECT perfusion study was performed with 99mTc and 201Tl using the same protocol that we use for patients. Baseline studies (with no defect inserted) were also performed. The SPECT images of the myocardial phantom with defects were compared with baseline SPECT images to determine whether the defect could be clearly identified.
Results:The uniformity of the baseline SPECT images was analyzed very carefully where an IP artifact was detected. The Dmin was determined for 99mTc and 201Tl at 3 radii of rotation: 21.0, 25.0, and 29.2 cm.
Conclusion:To be detected on SPECT images, a defect must be of a thickness ≥Dmin. A simple method for performing a quality control test for SPECT nuclear cardiology can be developed based on these findings.
- SPECT
- baseline SPECT images
- minimum detectable thickness
- external quality control test
- inferoposterior artifact
The left ventricle (LV) myocardial wall thickness ranges from 8 to 15 mm. To our knowledge, there has not been a conclusive study to assess the minimum perfusion defect thickness that can be detected by a SPECT test with 99mTc and 201Tl. For follow-up patients, it becomes important to detect and assess a myocardial perfusion defect not involving the full LV wall thickness. In this cardiac phantom study, we determined the minimum detectable defect thickness at different locations within the myocardial wall.
Materials and Methods
Cardiac Phantom
A cylindrical cardiac phantom (Data Spectrum Corp., Hillsborough, NC) with inner chamber diameter of 36 mm and length of 74 mm and outer chamber diameter of 60 mm and length of 86 mm was used to acquire the SPECT projection data. The space between the two chambers simulated the myocardial wall of the left ventricle (MWLV) and had a thickness of 12 mm and volume of 120 mL. The alignment pins, which were originally mounted on the outer wall of the inner chamber, were removed. Plastic defects were fixed on the outer wall of the inner chamber using a thin layer of Bulldog Grip glue (Drummond American Corp., Vernon Hills, IL), which worked perfectly in water solutions and was easy to remove. For SPECT studies the MWLV portion of the phantom was filled with 11.1 MBq (0.3 mCi) 99mTc water solution (1% of the patient dose) (1) or 4.44 MBq (0.12 mCi) of 201Tl solution (4% of the patient dose) (2). To ensure uniform distribution, the tracer was mixed with water before filling.
Quality Assurance
The study was performed using a single-head gamma camera system (Genesys; ADAC, Milpitas, CA) with a low-energy, high-resolution (LEHR) parallel hole collimator. To increase the clinical usefulness of SPECT studies, we performed the following quality control tasks:
The camera was carefully peaked every working day at energy level of 140 keV (20% window) for 99mTc and at energy levels of 72 keV (25%), 135 keV (15%), and 167 keV (20%) before 201Tl studies.
The intrinsic uniformity was measured daily using a point 99mTc source set at a distance of 310 cm from the center of the crystal, and the variation was always <2%.
The intrinsic spatial resolution was checked weekly using a 4-quadrant bar phantom and was always better than 3.2 mm.
The extrinsic spatial resolution of the system was evaluated by scanning a point 99mTc source (3) and using the count profile software program provided by the manufacturer. The point spread function (PSF) of the source (count density vs. pixel number) was obtained by generating a straight line through the source image. The extrinsic spatial resolution was determined by calculating the full width at half maximum (FWHM). The extrinsic spatial resolution was found to be 7.5 mm at the collimator surface and 8.7, 9.9, 11.6, and 13.9 mm at distances of 10, 21.0, 25.0, and 29.2 cm from the collimator, respectively.
The pixel size for the 64 × 64 matrix size and 38 × 38 cm2 field of view, which is routinely used to acquire the cardiac SPECT studies, was measured weekly to be 5.8 ± 0.2 mm.
The axis of rotation was checked weekly using the software program provided by the manufacturer. The error range for both the x-axis and the y-axis was always less than 2 mm for the 64 × 64 matrix size and full field of view.
Acquiring and Processing
The phantom was scanned using the same protocol that is used for imaging a myocardial SPECT patient. Phantom was placed on the imaging table approximately at the center of gantry rotation so that the long axis made an upward angle of 10–15° with the table surface. No attenuating media was placed around the phantom. The projection data were acquired in 64 × 64 matrix for a circular 180° orbit from the left posterior oblique (135°) position to right anterior oblique (−45°) position in a step-and-shoot mode. Sixty-four images (20 s per image) were acquired for 99mTc, and 32 images (40 s per image) for 201Tl studies. The raw data was processed by the AutoSPECT+ software program (ADAC, Milpitas, CA) using a gaussian filter (order, 5; cutoff frequency, 0.5 cycle/pixel). No attenuation or scatter correction was performed.
Baseline SPECT Images
The baseline SPECT phantom images were acquired for each radius of rotation (ROR) without any defects (inserts). These data were then used to compare the baseline images to the images with the defects inserted in the phantom. The baseline SPECT phantom images were also acquired using a 360° acquisition (64 images, 20 s per image) for circular (25.0-cm ROR) as well as elliptical (25.0-cm long-axis and 15.0-cm short-axis) orbits.
To estimate the degree of nonuniformity in baseline SPECT images, a bull’s-eye map was generated and the region ratio software program provided by the manufacturer was used to determine the average ratio of count density for the anterior-to-inferoposterior (ANT/IP) regions as well as for the septal-to-lateral (SEP/LAT) regions.
Determination of Minimum Detectable Defect Thickness
For experimental determination of minimum detectable defect thickness in phantom SPECT studies, we performed the following steps:
We inserted 4 plastic defects (cylindrical or rectangular) of 50-mm length, 20-mm width, and a thickness (depth) of 7 mm, so that the long-axis of the strip was parallel to the phantom long-axis. These strips were fixed at four regions—anterior (ANT), septal (SEP), inferoposterial (IP), and lateral (LAT)—on the inner wall of the MWLV portion of the phantom. A fifth defect of 20 × 20 × 7 mm3 was inserted to the apical (AP) region of the phantom. Figure 1 shows the defect locations inside the phantom. Note that all defects had the same thickness (or depth) for a given set of measurements.
The MWLV portion of the phantom was then filled with 99mTc water solution, and at three ROR (21.0, 25.0, and 29.2 cm) the SPECT studies were performed.
The phantom was refilled with 201Tl water solution and step 2 was repeated.
The thickness of the 5 defects was decreased by increments of 1 mm, and steps 2 and 3 were repeated for every thickness.
Locations of defects inside myocardial phantom. D = thickness (depth) of the defect.
Results
Baseline SPECT Images
The tracer distribution within the baseline SPECT images was nonuniform, especially for 201Tl. Figure 2 clearly shows this nonuniformity on short-axis, vertical long-axis, and horizontal long-axis baseline images with 201Tl for 21.0-cm ROR, where a clear loss of count density is seen in the IP region near the base, which is an artifact. Table 1 shows the average density ratios ANT/IP and SEP/LAT for 99mTc and 201Tl at three different ROR.
Average Density Ratios for Baseline SPECT Images
Example of short-axis, vertical long-axis, and horizontal long-axis slices showing nonuniformity of baseline SPECT images.
It was also found that the nonuniformity of the baseline SPECT images was independent of the arc of the acquisition orbit (180° or 360°), the orbit type (circular or elliptical), and the direction of rotation (clockwise or anticlockwise).
To verify that the IP defect was not specific to our camera, the same phantom was scanned without any inserted defects for 201Tl for ROR = 21.0 cm using 2 other gamma camera systems: single-head ADAC Genysys (DeTar Hospital, Victoria, TX) and dual-head Sopha DSTXL (Citizens Medical Center, Victoria, TX), and in both cases the IP artifact with 201Tl was detected.
Minimum Detectable Defect Thickness
Figure 3 shows the phantom SPECT images with 99mTc (row A) and 201Tl (row B) at 29.2-cm ROR, where 5 defects with a thickness of D = 6 mm were inserted at the 5 locations described in Figure 1. This figure shows that the SPECT images with 99mTc were sharper than those with 201Tl and that all defects were clearly seen with both tracers, except the defect located in the septal region, where it is seen on short-axis slices and barely seen on horizontal long-axis slices.
Phantom SPECT images for ROR = 29.2 cm, with 99mTc (row A) and 201Tl (row B) where 5 defects with thickness D = 6 mm were inserted. All defects are clearly detected with both tracers except defect located in SEP region in 201Tl study.
Table 2 summarizes the minimum detectable defect thickness (Dmin) in SPECT images for 99mTc and 201Tl as determined at 3 different ROR. Dmin is expressed in millimeters and as a percentage of the phantom myocardial wall thickness. Dmin is the minimum thickness of the defect at which the defect can still clearly be seen on 2 different projections of SPECT slices, and below which the defect cannot be clearly detected. From Table 2 we notice:
For the same location and same ROR, Dmin is less with 99mTc than with 201Tl.
Dmin is smallest when the defect is located in the IP region and greatest when the defect is located in the SEP region.
The phantom was scanned again to confirm the later result. The two 6-mL (40 × 50 × 3 mm3) defects were placed in the phantom in two opposing locations (ANT and IP), and rotating the phantom by 90°, the locations were changed to SEP and LAT. The SPECT images with both tracers at 25.0-cm ROR showed that the defects located in the IP region were the only ones that were clearly seen.
Minimum Detectable Defect Thickness for 99mTc and 201Tl
Figure 4 shows the phantom SPECT images with 99mTc and 201Tl at 25.0-cm ROR, where approximately 2-mL defects (22.5 × 22.5 × 4 mm3) were located in the ANT, SEP, IP, and LAT regions midway between the apex and the base. All 4 defects were detected with 99mTc, whereas only the defect located in the IP regions was clearly seen with 201Tl.
Phantom SPECT images with 99mTc (row A) and 201Tl (row B) at ROR = 25.0 cm. Four defects of 22.5 × 22.5 × 4 mm3 (∼ 2 mL) were fixed at ANT, SEP, IP, and LAT regions. All defects were detected with 99mTc; only the defect located in IP region was detected with 201Tl.
Figure 5 shows the phantom SPECT images with 99mTc at 25.0 cm ROR, where 2 defects of approximately the same volume (1 mL) but of different thicknesses (6 vs. 3 mm) were located at the ANT, SEP, IP, and LAT regions (2 defects for each region). The first defects of 13 × 13 × 6 mm3 were located toward the apex, and the second defects of 18.5 × 18.5 × 3 mm3 were located toward the base. The separation between the 2 defects was 10 mm. Row B on Figure 5 shows the phantom baseline SPECT images with 99mTc for 25.0-cm ROR. In Figure 5, the 6-mm-thick defects are clearly seen for all 4 regions, whereas the 3-mm-thick defects near the base are clearly visible only for the IP region.
Phantom SPECT images for ROR = 25.0 cm with 99mTc (row A). Two 1-mL defects were inserted into ANT, SEP, IP, and LAT regions. All 6-mm defects are clearly seen; 3-mm defects are visible in IP region only.
Discussion
The loss of counts in the inferior wall of SPECT images where no defect exists has been reported in phantom (4,5) and clinical (6) studies. Lowe et al. (4), in a thoracic phantom study that detected this artifact with 201Tl, noticed that the defect margins in the inferior wall were less distinct in 201Tl SPECT images when attenuating media were present. Ye et al. (5) performed an anthropomorphic phantom study with 99mTc and 201Tl that used an attenuation correction and resolution recovery software program to correct for this artifact. They found that the uniformity of SPECT images in regions where no defect presented was improved in 99mTc studies, while remaining unchanged in 201Tl studies.
Our results showed that the uniformity of phantom baseline SPECT images improved as ROR increased (Table 1) and that the appearance of the IP artifact was independent of orbit type, orbit arc, direction of rotation, and gamma camera type. These facts may indicate that this artifact is related to physical factors associated with imaging the phantom, such as:
Partial volume effect, since the phantom wall thickness is of a similar dimension to the spatial resolution of the gamma camera;
Degradation of spatial resolution as object-to-camera distance increases; or
Nonlinear attenuation of γ-rays through the phantom material.
All the preceding factors may produce regional reduction in the perfusion of baseline SPECT images, but do not explain the constant location of this artifact in the IP region.
Although cardiac diagnosis does not normally demand absolute quantification of defect size or severity, information about these quantities is important for the follow-up of cardiac patients.
The values of Dmin given in Table 2 cannot be extrapolated to real clinical studies because of the lack of cardiac and respiratory motion, absence of attenuation and scatter media, and the absence of gamma background from nearby organs in the phantom experiment.
Several software techniques have been developed to determine defect size and severity (4,7,8) using cardiac SPECT perfusion images. O’Connor et al. (7) in their multicenter study with cardiac phantoms, analyzed 5 representative short-axis slices of 99mTc perfusion SPECT images to determine defect size and severity. They used 8 defects of relatively large volumes (5%–70% of the myocardial phantom volume) and reported that the error in defect size determination was <2%. Benoit et al. (8) used a 360° elliptical sampling of radial slices to evaluate defect size and severity in their Data Spectrum phantom study and in clinical studies. In the phantom study they used 7 defects with volumes of 6–48 mL to evaluate the software performance, and the defect size appeared to be accurately estimated, with mean relative standard deviation (RSD) of <8%. Lowe et al. (4), in their thoracic phantoms study to evaluate the dual-isotope (201Tl rest/99mTc-sestamibi stress) technique, used the bull’s-eye program to determine perfusion defect size and severity where 3 defects with volumes of 4.9, 9.5, and 19.2 mL were used. In this study the absolute error in determining the defect size for a single agent (99mTc or 201Tl) increased as volume size decreased (for 99mTc without attenuation media, the absolute error was 8% for the 19.2-mL defect and 37% for the 4.9-mL defect). Our results showed that the appearance of small defects (1–4 mL) in SPECT images depends on defect thickness, defect location, acquisition ROR, and tracer type (99mTc or 201Tl), especially when the defect thickness approaches Dmin. A correction factor may have to be introduced in software programs designed to determine defect size and severity.
Our experiment showed that Dmin is smallest (easiest to detect) for defects located in the IP region and greatest for defects located in the SEP region. This finding correlates well with the clinical results of Meyers et al. (9), in which the specific locations of cardiac wall abnormalities were assessed in a population of 562 patients with CAD using the dual-isotope perfusion (201Tl rest/99mTc-sestamibi stress) test. The authors found that the largest percentage (35%) of defects were recorded for the inferior wall, whereas only 13.8% were found on the septal wall.
Heikkinen et al. (10) reported the results of a multicenter study to evaluate the quality of cardiac perfusion SPECT procedures in Finland that showed the need for an external quality control test in nuclear cardiology (in addition to the traditional SPECT QC test of uniformity, relative sensitivity, axis of rotation, pixel calibration, and spatial resolution). They also recommended the use of a LEHR collimator, 4–6 mm pixel size, the location of the myocardium closer to the center of rotation (COR), and the use of corrections for COR, uniformity, and attenuation.
Approximately 70%–75% of our workload in the nuclear medicine department is cardiac perfusion SPECT with 99mTc cardiolite and 201Tl. For the external or overall QC test we suggest the configuration of defects shown on Figure 1, with a defect thickness of 6 mm. The phantom wall cavity is filled with 11.1 MBq of 99mTc water solution (or 4.44 MBq of 201Tl solution) and the SPECT data are acquired at 25.0-cm ROR with 10 s per step for 99mTc and 20 s per step for 201Tl. All defects were clearly seen on the reconstructed SPECT images using this procedure. This QC procedure takes 11 min to perform before acquisition of our clinical cardiac studies. Any deviation in the appearance of SPECT images may indicate a problem that should be corrected before acquiring patient data.
Conclusion
This study showed that defect thickness and location are important factors determining whether a defect would be detected in phantom myocardial perfusion studies. For a defect to be detected, it must have a threshold value ≥Dmin, which depends on the defect location within the myocardial wall, the acquisition ROR, and the tracer type (99mTc or 201Tl). It was planned to use this phantom as a part of our QC procedure. The primary benefit obtained from this study was the determination of the best defect thickness for performing the QC procedure. This is a simple method for rapidly performing a QC test that is directly applicable to nuclear cardiology.
Acknowledgments
The authors thank the administration of Cuero Community Hospital, Cuero, TX, for supporting the research in all its stages. We thank Professor Hussein M. Abdel-Dayem, MD; Steven C. Schnicker, MD; Stephen W. Tibbits, MD; James F. Neuman, MD; and Ajay K. Gaalla, MD, for fruitful discussions; and J. D. Davis and Judy Heil for help with manuscript preparation.
Footnotes
For correspondence or reprints contact: Abdelhamid Elkamhawy, Nuclear Medicine Supervisor, Cuero Community Hospital, 2550 N. Esplanade, Cuero, TX 77954; Phone: 361-275-6191 (ext. 476); Fax: 361-275-9908; E-mail: abdel{at}dewittec.net.