B Academy of Molecular Imaging, 2006 Published Online: 12 October 2006
Mol Imaging Biol (2006) 8:324Y332 DOI: 10.1007/s11307-006-0058-z
RESEARCH ARTICLE
In Vivo Quantitation of Intratumoral Radioisotope Uptake Using Micro-Single Photon Emission Computed Tomography/Computed Tomography Stephanie K. Carlson, MD, MS,1,3 Kelly L. Classic, MS,2 Elizabeth M. Hadac, BS,3 Claire E. Bender, MD,1 Bradley J. Kemp, PhD,1 Val J. Lowe, MD,1 Tanya L. Hoskin, MS,4 Stephen J. Russell, MD, PhD3 1
Department of Radiology, Mayo Clinic, 200 First Street SW, Rochester, MN 559005, USA Section of Safety, Mayo Clinic, 200 First Street SW, Rochester, MN 559005, USA 3 Molecular Medicine Program, Mayo Clinic, 200 First Street SW, Rochester, MN 559005, USA 4 Division of Biostatistics, Mayo Clinic, 200 First Street SW, Rochester, MN 559005, USA 2
Abstract Purpose: This study was undertaken to determine the ability of micro-single photon emission computed tomography (micro-SPECT)/computed tomography (CT) to accurately quantitate intratumoral radioisotope uptake in vivo and to compare these measurements with planar imaging and micro-SPECT imaging alone. Procedures: Human pancreatic cancer xenografts were established in 10 mice. Intratumoral radioisotope uptake was achieved via intratumoral injection of an attenuated measles virus vector expressing the NIS gene (MV-NIS). On various days after MV-NIS injection, 123I planar and micro-SPECT/CT imaging was performed. Tumor activity was determined by dose calibrator measurements and region-of-interest (ROI) image analysis. Agreement and reproducibility of tumor activity measurements were assessed by BlandYAltman plots and Lin_s concordance correlation coefficient (CCC). Results: Intratumoral radioisotope uptake was detected in all mice. Scatterplots demonstrate strong agreement (CCC=0.93) between micro-SPECT/CT ROI image analysis and dose calibrator tumor activity measurements. The differences between dose calibrator activity measurements and those obtained with ROI image analysis of micro-SPECT alone and planar imaging are less accurate and more variable (CCC=0.84 and 0.78, respectively). Conclusions: Micro-SPECT/CT can be used to accurately quantify intratumoral radioisotope uptake in vivo and is more reliable than planar or micro-SPECT imaging alone. Key words: Molecular imaging, Quantitation, SPECT, Micro-SPECT, Mice, Sodium iodide symporter (NIS) gene, Measles virus, Reporter gene
Introduction
R
adioactive tracers have a major role in preclinical molecular imaging of small animals [1Y6]. Prior to the
Correspondence to: Stephanie K. Carlson, MD, MS; e-mail:
[email protected]
development of dedicated small animal imaging systems, most preclinical animal imaging using radioisotopes was performed with planar or single photon emission computed tomography (SPECT) techniques using clinical systems. Planar imaging has been useful because it can demonstrate whole-body biodistribution of the injected radioisotope and
S.K. Carlson et al.: Small Animal Micro-SPECT/CT Quantitation it facilitates imaging of multiple small animals at one time. However, it has considerable limitations because of its low resolution and because it represents a 3-D distribution of radioactivity in a 2-D display [1, 7, 8]. This results in superimposition of overlying radioactivity, difficulty with accurate localization of abnormalities, and errors in radiotracer quantitation [9]. Dedicated small animal nuclear medicine-based imaging systems are being developed and used with increasing frequency in preclinical gene therapy research protocols to monitor gene expression and therapeutic efficacy [1, 2, 5, 10]. Combined SPECT/computed tomography (CT) scanners have been recently developed that allow fusion of functional (SPECT) data with high-resolution anatomical (CT) data to provide a more accurate localization and quantitative estimate of radioactivity in live animals [1, 11, 12]. The ability to use these systems for accurate quantitation of in vivo intratumoral gene expression and associated radioisotope uptake in living animal models and to monitor the effects of therapy over time is crucial for the translation of many preclinical studies to human clinical trials. Advances in quantitation will also enable comparison of study results between investigators and improve tumor dosimetry calculations to ensure maximal therapeutic effect with minimal toxicity to normal tissues. The purpose of this study was to determine the ability of micro-SPECT/CT to accurately quantitate intratumoral radioisotope uptake in vivo and to compare these measurements with those obtained by planar imaging and microSPECT imaging alone.
Materials and Methods In Vitro Assessment of True Tissue Radioactivity Using a National Institute of Standards and Technology-Calibrated Dose Calibrator National Institute of Standards and Technology (NIST)-calibrated dose calibrators are well-type chambers considered to be the reference standard for accurate quantitation of radioactivity greater than 3.710j4 MBq. Because they are highly susceptible to errors due to geometric variation and spatial location of the source within the chamber [13], we needed to validate the ability of our dose calibrator to accurately measure intratumoral radioisotope uptake. Cubes of Spam measuring 1.0, 1.5, and 2.0 cm3 simulated in vivo soft tissue tumors of similar size and similar photon attenuation characteristics. Approximately 7.4 MBq of iodine123 (123I) was injected into the center of each of the three different-sized cubes. Activity in the syringes was measured before and after injection to accurately assess the injected activity. The 1.0-cm3 cube containing 7.18 MBq of 123I (as determined by preinjection and postinjection syringe activity measurement) was used to determine if there was spatial dependence. The dose calibrator chamber is tubular [6 cm (diam) 25 cm (depth)] and surrounded by sealed argon gas-filled walls. Activity readings
325
were taken starting with the tissue cube at the bottom center of the chamber and moving it through the tube in 1-cm increments. To check for geometric variation between activity measurement in a plastic syringe and in tissue, each of the cubes was placed into the dose calibrator and measured activity was recorded by using a manufacturer-specified dial-in setting of 277 for 123I. These results were compared to actual injected activity, and a conversion factor for 123I activity measurement in tissue versus syringe was calculated for use with the actual mouse tumor activity measurements. All activity measurements were corrected for decay (123I half-life = 13.27 hours).
Micro-SPECT/CT Imaging System Animal imaging was performed by using a high-resolution microSPECT/CT system (X-SPECT, Gamma Medica-Ideas, Inc., Northridge, CA, USA). A low-energy, high-resolution parallel-hole collimator with a 12.5-cm field of view was used in all cases. Image acquisition time was five minutes for planar and 13 minutes for micro-SPECT imaging (64 projections at 10 seconds per projection). Micro-CT image acquisition (155-mm slice thickness, 256 images) was performed in one minute at 0.25 mA and 80 kVp. X-SPECT enables the animal to remain stationary on the same platform during imaging while the X-ray tube and gamma cameras share a common gantry and rotate around the bed. This allows highly accurate fusion and coregistration of the CT and SPECT data. Coregistration of the SPECT and CT images was performed by applying precalibrated spatial transformation to the SPECT images to match with the CT field of view. The CT and SPECT systems were used with a fixed imaging geometry throughout the course of the study (X-ray tube focus-to-isocenter distance of 225 mm; X-ray tube focus-to-detector distance of 289 mm). The precision of coregistration was tested by the manufacturers of the X-SPECT system (Gamma Medica-Ideas, Inc.) before and immediately after installation of the system at our institution by acquiring CT and SPECT images of capillary tubes filled with Tc-99m solution, coregistering these images, and measuring the differences between the center of the capillary tubes on the CT and SPECT images. This test resulted in a coregistration precision of 1 mm. CT images were reconstructed by using a modified Feldkamp (cone-beam filtered backprojection) reconstruction algorithm into a 5123 matrix with a voxel size of 0.1557 mm. This results in a field of view of 7.97 cm (diam) 7.97 cm (length). The CT system is capable of producing images with a resolution of 50 mm; however, with this reconstruction setting the voxel size limits the CT resolution to 0.3114 mm. The SPECT images were reconstructed by using a Filtered Backprojection algorithm into a 803 matrix size with a voxel size of 1.5 mm. This resulted in a field of view of 120 mm (diam) 120 mm (length). The resolution of the SPECT system is limited by the collimator resolution with parallel-hole collimators and is approximately 3Y4 mm.
Attenuated Measles Virus Expressing the Sodium Iodide Symporter Reporter Gene A recombinant attenuated measles virus vector expressing the sodium iodide symporter (MV-NIS) gene was used in this study to deliver the NIS gene to the target tumor. MV-NIS infection of tumor xenografts leads to tumor cell transduction and significant
326
S.K. Carlson et al.: Small Animal Micro-SPECT/CT Quantitation
intratumoral NIS expression. The utility of NIS as a molecular imaging reporter gene has been well demonstrated [14, 15]. Transfer of NIS to nonthyroidal cells induces iodide uptake activity similar to or greater than that seen in the thyroid [14, 15].
In Vivo Imaging Studies All studies involving animals were approved by the Institutional Animal Care and Use Committee. Ten six-week-old female nude mice (Harlan SpragueYDawley, Madison, WI, USA) were injected subcutaneously in the right flank with BXPC-3 human pancreatic cancer cells (107 cells/100 mL phosphate buffered saline). On day 9 after tumor cell implantation (median tumor size 7 mm), intratumoral (i.t.) injection of MV-NIS was performed by using 3.5 106 TCID50/100 mL Opti-MEM and a 25-gauge needle. The mice were imaged on days 2, 3, 5, 8, and 10 after MV-NIS infection (n = 2 mice per day). These times were chosen from previous data (not shown) regarding optimal MV-NIS expression in other human xenograft models. Planar and micro-SPECT/CT image acquisitions were obtained for all mice 1 hour after receiving an intraperitoneal injection of 18.5 MBq of 123I (corrected for decay and postinjection syringe measured activity). Mice were anesthetized for imaging with a 20-ml mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg). On each of the five imaging days, the two mice were euthanized by an anesthetic overdose immediately after imaging. Whole body activity (injected dose) in each of the euthanized mice was determined by measuring activity in the syringe in a NIST-calibrated dose calibrator before and after injection. Flank tumor activity was determined by two independent methods. First, flank tumors were excised and weighed, and radioisotope uptake was measured in the dose calibrator. Independently, region of interest (ROI) analysis of the planar, micro-SPECT alone, and micro-SPECT/CT fused images was performed and accumulated intratumoral counts were recorded. The two separate methods of measurement were then evaluated to determine the accuracy of the ROI analysis methodology compared with the reference standard dose calibrator measurements.
Planar and Micro-SPECT/CT Imaging of Standard
123
I
To determine the conversion factor for calculating in vivo intratumoral radioisotope uptake from ROI count measurements obtained with our image analysis software, planar and microSPECT/CT scanning of an 123I standard measuring approximately 18 mm3 and containing 1.77 MBq was performed with the same image variables used during in vivo mice imaging. An ROI was drawn for the standard on both the planar and micro-SPECT/CT images, and accumulated counts were determined. The size of the ROI was determined by tracing around all visualized activity (excluding background activity) on the planar image and by using the CT image to trace the ROI outline on the micro-SPECT/CT image acquisition. These correction factors (standard ROI counts/ standard activity) were used to calculate tumor activity on the planar, micro-SPECT alone, and micro-SPECT/CT images.
Image Analysis PMOD Biomedical Image Quantification and Kinetic Modeling Software (PMOD Technologies, Switzerland) was used for image analysis. All planar, micro-SPECT alone, and fused SPECT/CT images were adjusted for equal image intensity. On the micro-SPECT/CT images, the axial CT images were used to define the ROI around the tumor margin on every CT image in which it was visualized, resulting in a volume of interest (VOI). Pixel counts within the tumor regions were measured from the coregistered micro-SPECT images. Using the same image intensity for the micro-SPECT alone and planar images, ROIs were drawn around the tumor regions [multiple ROIs (VOI) for micro-SPECT and single ROI for planar projection] by using a subjective visual tumor boundary method in which attempts were made to draw the ROI boundary as close as possible to the presumed tumor boundary. ROIs were drawn with the goal of excluding adjacent normal structures (e.g., pelvic bones, bladder). Corresponding total intratumoral pixel counts were converted to activity by using the equations derived from scanning an 123I standard containing a known amount of radioactivity (described above). Background uptake was measured and corrected for by ROI image analysis of the normal opposite flank tissue and subtracted from intratumoral activity measurements obtained by ROI image analysis of the tumor uptake.
Image Analysis Using a 50% Threshold Edge Detection Technique To improve quantitation on the basis of the SPECT alone images, ROI analysis was performed by using a threshold edge detection technique. This has been described in the literature, is used routinely in clinical nuclear medicine, and is considered to be more accurate for quantitation than measurements based on the visual boundary method [16]. In clinical nuclear medicine, thresholds ranging from 25% to 75% are used. For this study, we chose to use a 50% threshold. The isocontour (or count level) was set at 50% of the maximum count in the tumor and the ROI was drawn around the corresponding area of maximum activity in the tumor on every micro-SPECT slice in which tumor was visualized.
Determination of Recovery Coefficient Correction Factors Calculated intratumoral radioisotope activity can be underestimated because of partial volume effects intrinsic to the imaging system. The recovery coefficient, or the ratio of measured activity concentration to actual activity concentration, serves as a correction factor for this partial volume effect and was determined for our system by imaging sources of various known sizes and known activity concentrations. Spherical phantoms ranging in inner diameter from 4.95 to 15.43 mm, each loaded with 9,435 MBq/ mL 123I, were imaged along with a standard known to be much larger than the system full-width, half maximum (FWHM). The standard was used to determine the conversion from count concentration to activity concentration, which was then applied to each sphere measurement. For each mouse in our study, the recovery coefficient was determined based on tumor volume (as
S.K. Carlson et al.: Small Animal Micro-SPECT/CT Quantitation determined on ROI analysis of the micro-SPECT/CT images) and was applied to the previously calculated data to determine its impact on our initial calculations.
Statistical Analysis Measurements of iodide uptake and calculated tumor activity by in vivo micro-SPECT/CT, micro-SPECT alone, and planar image analysis were compared to the true values, as measured ex vivo with the NIST-calibrated dose calibrator. The agreement of in vivo measurements with ex vivo measurements was assessed by BlandYAltman methodology [17]. For each agreement analysis the following plots were created: (1) in vivo vs. ex vivo measurement with line of equality and (2) ex vivo minus in vivo difference versus the mean of the two. We also calculated the 95% limits of agreement. Lin_s concordance correlation coefficient (CCC) [18] was used to quantify the ability of micro-SPECT/CT, micro-SPECT alone, and planar imaging to reproduce the true iodide uptake. This statistic is similar to the Pearson correlation coefficient but specifically measures the variability about the equality line rather than a general linear relationship; two methods showing perfect agreement would have CCC = 1. Paired t tests were also used to compare in vivo measurements to the ex vivo measurement. P values less than 0.05 were considered statistically significant. Analysis was performed using SAS software (SAS Institute, Inc., Cary, NC, Version 8).
327
Planar and Micro-SPECT Tumor Activity Conversion Factors To establish the conversion factor for translating accumulated counts into activity, a 1.77-MBq 123I standard was imaged with planar (5 minutes dorsal projection) and microSPECT imaging (1,200 counts per image at 10 seconds per image for a total of 64 images). The planar image corresponded to 25,797 counts, resulting in a conversion factor of 6.88 10j5 MBq/measured count. For microSPECT, this activity corresponded to 1.10 107 counts, resulting in a conversion factor of 1.63 10j7 MBq/ measured count.
Measurement of Intratumoral Radioisotope Uptake On the micro-SPECT/CT images, the axial CT images were used to define the ROI around the tumor margin and
Results Dose Calibrator Measurements No vertical spatial dependence was seen in the measurements of activity for the dose calibrator and tissue cubes Ve.g., measurements of the tissue at the bottom of the chamber and midchamber achieved the same result. Measurements of tissue activity compared with plastic syringe activity resulted in a conversion factor of 1.11, suggesting that there was some minor attenuation of the 123I photons within the tissue. Although this varied slightly (1.10Y1.12) with the size of the tissue cube, we chose to use only the value 1.11. This factor was multiplied by the tumor activity measurement to obtain a corrected tumor activity for comparison with ROI analysis results.
Imaging of Radioisotope Uptake with Planar and Micro-SPECT/CT Imaging Planar and micro-SPECT/CT imaging documented intratumoral radioisotope uptake at varying levels in all 10 mice on planar and fusion micro-SPECT/CT images (Fig. 1). The background activity in the control flank was negligible, as determined by dose calibrator measurements and ROI analysis.
Fig. 1. Images of mice with intratumoral 123I uptake (arrows) on planar (left) and micro-SPECT/CT (right). Note the other physiologic areas of increased uptake in the thyroid gland and bladder (planar image) and stomach (planar and microSPECT/CT images). The background activity in the control flank was negligible, as determined by dose calibrator measurements and ROI analysis.
328
S.K. Carlson et al.: Small Animal Micro-SPECT/CT Quantitation
corresponding pixel counts were measured from the coregistered micro-SPECT images (Fig. 2). Injected dose and the intratumoral 123I uptake as determined by dose calibrator measurements and ROI image analysis calculations for each of the three imaging modalities are shown in Table 1. BlandYAltman analysis of agreement scatterplots (Fig. 3) with associated Lin_s CCCs shows that there is strong agreement between microSPECT/CT and dose calibrator tumor activity measurements. The difference between dose calibrator radioactivity measurements and ROI image analysis measurements of micro-SPECT alone and planar imaging are much more
variable and less accurate (CCC = 0.84 and 0.78, respectively) than those for micro-SPECT/CT. In some cases, the micro-SPECT alone and planar measurements are twice the true measured tumor activity. The scatterplots and paired t tests demonstrate a systematic measurement bias with micro-SPECT/CT consistently underestimating tumor activity (P = 0.003) and planar imaging consistently overestimating tumor activity (P = 0.01).
Activity Calculations Using a 50% Threshold Edge Detection Technique for Micro-SPECT Alone Images Results are also shown in Table 1 for 123I intratumoral uptake determined by ROI analysis of micro-SPECT images alone by using a 50% threshold edge detection technique compared with the original micro-SPECT ROI analysis technique. BlandYAltman analysis of the micro-SPECT alone measurements versus dose calibrator measurements (scatterplots not included) show that use of the 50% threshold edge detection technique does not improve agreement with the true intratumoral dose calibrator measurements (CCC = 0.66); however, it does decrease the systematic bias of consistent overestimation of tumor size and activity. Paired t test results showed a statistically significant bias (P = 0.01) using the original visual boundary method compared with no statistically significant systematic bias (P = 0.16) using micro-SPECT alone with the 50% threshold edge detection technique.
Impact of Recovery Coefficient Correction Factors on Intratumoral 123I Uptake Measurements
Fig. 2. Example of ROI analysis technique on microSPECT/CT. Axial SPECT/CT images of a mouse flank demonstrate the ability to accurately quantitate intratumoral radioisotope uptake by using the anatomical CT image and the corresponding coregistered SPECT data to precisely define the tumor ROI (yellow squares with connecting blue line = outline of flank tumor). Measured tumor activity in this mouse using ROI analysis = 0.70 MBq. Ex vivo dose calibrator measurements of true tumor activity = 0.89 MBq.
Recovery coefficient correction factors for tumor activity in the flank region were applied to the original micro-SPECT/ CT, micro-SPECT alone, and micro-SPECT alone using a 50% threshold edge detection technique data sets to determine the impact of these correction factors on intratumoral activity measurements. Results of our recovery coefficient experiment were plotted as a function of object (or tumor) volume to facilitate use with small animal studies (Fig. 4). The recovery coefficient correction factor was applied by dividing the intratumoral 123I uptake determined by ROI analysis with the appropriate recovery coefficient based on tumor volume (as measured on ROI image analysis). The revised calculated intratumoral activities with the recovery coefficient applied are listed in Table 1. Results from micro-SPECT/CT ROI analysis continue to show strong agreement (CCC = 0.94) with dose calibrator measurements, whereas micro-SPECT alone shows a slightly weaker agreement (CCC = 0.82) (Fig. 5). The corrected micro-SPECT/CT versus dose calibrator scatterplot and paired t test also show a smaller mean difference between
S.K. Carlson et al.: Small Animal Micro-SPECT/CT Quantitation
329
Table 1. 123I injected dose and intratumoral radioisotope uptake measured by dose calibrator and ROI image analysis using micro-SPECT/CT, microSPECT alone, and planar imaginga with/without application of recovery coefficient correction factors Mouse
Injected activity
Tumor weight (g)
DCb
ROI analysis of tumor activity No correction factors applied
1 2 3 4 5 6 7 8 9 10
18.1 17.6 15.5 15.7 18.0 17.6 18.6 17.8 18.6 16.4
0.403 0.692 0.225 0.888 1.400 0.766 1.250 0.600 0.435 1.291
0.89 1.45 0.31 1.16 1.81 0.86 1.25 0.55 0.39 0.81
Recovery coefficient applied
MicroSPECT/CT
Micro-SPECT alone
Micro-SPECT alone with 50% threshold
Planar
MicroSPECT/CT
Micro-SPECT alone
Micro-SPECT alone using 50% threshold
0.70 1.30 0.29 1.05 1.48 0.68 1.10 0.58 0.29 0.73
1.33 1.84 0.64 1.09 1.85 0.63 1.64 0.69 0.51 0.49
0.85 1.37 0.44 0.84 1.05 1.01 1.10 0.97 0.49 0.78
1.30 1.66 0.62 1.17 1.81 0.97 1.84 0.99 0.75 0.71
1.08 1.38 0.39 1.09 1.48 0.73 1.29 0.65 0.51 0.74
1.41 1.84 0.64 1.13 1.85 0.71 1.64 0.76 0.62 0.58
1.16 1.45 0.49 0.95 1.18 1.01 1.29 0.97 0.62 0.78
a
All results are expressed in MBq. All results corrected for radioisotope decay (123I half-life = 13.2 h). Dose calibrator (reference standard).
b
micro-SPECT/CT and dose calibrator measurements. This comparison is no longer statistically significant (P = 0.80), suggesting that the correction helps remove the systematic bias observed in the uncorrected micro-SPECT/CT measurements.
Discussion We have shown that NIS-mediated intratumoral radioisotope uptake can be accurately monitored noninvasively and quantitatively by in vivo imaging with micro-SPECT/CT. By using the anatomical micro-CT images to accurately define the tumor margins and the precisely coregistered micro-SPECT images to determine the associated intratumoral counts, we have demonstrated strong agreement
between in vivo micro-SPECT/CT ROI image analysis and true intratumoral radioisotope uptake measurements measured ex vivo. Noninvasive monitoring of radioisotope uptake and therapeutic response is critical for the advancement of preclinical radiotherapy trials into the clinic. In vivo microSPECT/CT can follow the time course of intratumoral radioisotope uptake and longitudinal studies can be performed with repeated measurements over time on the same animal. Optimal monitoring of radiotherapy effectiveness and early response requires quantitative rather than qualitative measures [6, 16, 19, 20]. Imaging protocols and quantitation methods need to be standardized, consistent, and specific for the radioisotope, imaging system, and image analysis
Fig. 3. BlandYAltman analysis of agreement plots with identity line added. Plots display dose calibrator activity measurements (x-axis) versus ROI analysis activity measurements (y-axis) for micro-SPECT/CT (A), micro-SPECT alone (B), and planar (C) imaging. Lin_s concordance correlation coefficient (CCC) measures the strength of the correlation about the identity line (diagonal solid line).
330
S.K. Carlson et al.: Small Animal Micro-SPECT/CT Quantitation
Fig. 4. Recovery coefficient graph determined by imaging 123I sources of various known sizes and activity concentrations.
software used. Accurate quantitation is particularly important for translation of patient-specific targeted radionuclide therapy protocols into human trials because correct dosing is needed to achieve a therapeutic antitumor effect and to minimize toxicity to normal structures [6, 8, 21, 22]. For clinical therapeutic applications, deviations of 20% or more can impact the actual effectiveness of the treatment (due to an underdose or overdose), can cause adverse biological effects in organs not being treated but which accumulate the radiopharmaceutical, and are considered reportable medical events by the U.S. Nuclear Regulatory Commission [23Y27]. More rapid assessment of therapeutic response will also allow clinical oncologists to make much earlier treatment decisions than in the past and shorten the expensive and time-consuming process involved with the development of new cancer gene therapies. Planar or SPECT imaging alone is suboptimal for intratumoral radioisotope quantitation because of low
spatial resolution and the inability to define the ROI accurately. Inaccurate ROI measurements can overestimate or underestimate the size of the tumor and its associated radioisotope uptake. Planar imaging also has significant limitations because it represents a 3-D distribution of radioactivity in a 2-D display. This results in superimposition of overlying radioactivity, difficulty with precise localization of abnormalities, and errors in quantitation. ROI measurements need to be standardized and to reflect the true tumor size to adequately assess changes in radionuclide uptake [6]. The apparent volume of an object changes with display intensity. Inappropriately increasing image gain or intensity does not represent an increase in uptake and can easily be misinterpreted as such. The addition of CT imaging allows precise coregistration of the CT and SPECT images and provides an anatomical outline. ROI determined by CT tumor structural boundaries and the associated intensity (increased or decreased to ensure that the counts are within those boundaries) allows accurate assessment of true tumor activity uptake. Although it is not always easy to distinguish tumor from soft tissue in CT, one can often see tumor mass effect or identify fat planes between the tumor and adjacent muscle or other soft tissue when defining ROIs on subcutaneous xenografts. Orthotopic xenografts, such as pancreatic or hepatic tumors, may be more difficult to distinguish from the adjacent organ or soft tissue. In these cases, using intravenous CT contrast may aid in the accurate placement of tumor ROIs. Although CT images can be used to draw accurate ROIs around the tumor or other targets of interest, we have found whole-body ROI measurements to be imprecise and subject to significant error. In addition, delays in imaging affect whole-body ROI measurements by ignoring radioactivity excreted in the urine and feces and by failing to take into account other sources of uptake on the image (e.g., skin contamination). A more definitive approach to determine
Fig. 5. BlandYAltman analysis of agreement plots displaying dose calibrator activity measurements (x-axis) versus ROI analysis activity measurements (y-axis) for micro-SPECT/CT and micro-SPECT alone after application of the recovery coefficient. Micro-SPECT/CT ROI analysis continues to show strong agreement with dose calibrator measurements, whereas micro-SPECT alone shows a slightly weaker agreement with dose calibrator measurements.
S.K. Carlson et al.: Small Animal Micro-SPECT/CT Quantitation percentage of injected dose in the tumor is to use the known true injected dose. This can be done by measuring the postinjection syringe activity in a dose calibrator, subtracting this amount from the preinjection syringe activity measurement, and correcting for time-related radioisotope decay. This methodology is still subject to error if there is infiltration and residual activity at the injection site. If infiltration does occur, this should be detected during imaging of the animal. With micro-SPECT/CT ROI analysis, we consistently underestimated tumor activity. Although our results support the idea that the ROI analysis is consistent, they suggest some systematic bias in the micro-SPECT/CT measurements. The limiting factor for accurate quantitation is spatial resolution [10]. As a result of system resolution, the activity recovered during an emission scan depends on the size of the imaged object. When the object is at least two to three times larger than the FWHM of the system, the intensity of the images is directly related to the amount and the concentration of the source. If the object is smaller than twice the FWHM, it then occupies less than one full voxel and, although the intensity of the image maintains a correlation with the total activity, the activity concentration is no longer accurately portrayed because of the distribution of the emitted activity spread out over a larger volume [1, 10]. The recovery coefficient is a correction factor for this partial-volume effect [28]. Because the recovery coefficient depends on object size, it is important to use accurate estimates of tumor volume [3]. Caliper measurements of tumor size are prone to error [29]. Accurate tumor volume determination by CT ROI analysis aids in the application of the recovery coefficient to calculated activity measurements. Radioactivity uptake calculations can also be affected by soft tissue attenuation of photons that need to travel through the mouse body to be detected. Photon attenuation using 123 I for micro-SPECT/CT imaging of small animals reportedly ranges from 3% and 15% in mouse-sized objects [10, 30, 31]. Attenuation coefficients can be used to correct for photons lost by attenuation. The software used to determine attenuation coefficients with CT attenuation maps and iterative reconstruction methods is not currently operational on our micro-SPECT/CT system, therefore we did not take into account attenuation correction factors in this study. Although attenuation is considered not as important in small animals, results for intratumoral radioisotope calculations may be more accurate when this is taken into account [9Y11, 32]. We also did not attempt to correct for photon scatter in this study because the amount of scatter is typically considered insignificant in mice owing to the shorter path lengths that photons travel [11]. All institutions do not have fusion micro-SPECT/CT capability. When only micro-SPECT imaging is available, a threshold edge detection technique can be used to improve radioisotope quantitation by decreasing the variability of ROI measurements [16]. This method is routinely used in clinical nuclear medicine practice and decreases the influence of
331
background image noise and corrects for adjacent areas of uptake that may spill over into the tumor but are not actually tumor uptake. In our study, we used a 50% threshold level in attempts to improve micro-SPECT ROI analysis. Although this method did decrease the variability of our measurements compared with the micro-SPECT visual boundary method, the results still were not as accurate as those obtained by fusion micro-SPECT/CT ROI analysis. Image misregistration can also contribute to a large error in calculating tumor dosimetry [22]. Our micro-SPECT/CT system allows for near-perfect coregistration owing to the ability of the SPECT and CT images to be acquired sequentially without removing the animal from the imaging system. Our system_s ability to perform accurate coregistration of the CT and SPECT images has been validated and therefore, misregistration is likely of minimal concern in our study. There are potential limitations to our study. Dose calibrators have a reported accuracy of T5% with 3.7 10j4 MBq (10 nCi) or more of radioactivity, and geometric variations can result in errors of up to 200% [13]. We reduced these errors by calibrating the dose calibrator used with an NIST-traceable standard and correcting for geometric inconsistencies between a plastic syringe and tissue. As a result, we estimate the error associated with measurement of activity in the dose calibrator to be less than 3%. There also is potential for observer bias in drawing the tumor ROI, especially for the planar and micro-SPECT alone images, with considerably less effect on the micro-SPECT/CT images because the anatomical CT images can be used as a guide for ROI placement. An additional study is underway to determine intra- and interobserver variability of tumor ROI placement and its effect on tumor uptake values. We did not attempt to quantify the amount of radioactivity excreted in the mice urine and feces. This, however, did not have an effect on our whole body calculations because we used the total injected dose to the mouse (from the preinjection and postinjection syringe activity counts as measured in the dose calibrator) rather than relying on whole-body ROI counts, which can be biased by loss of excreted radioactivity. Finally, our system uses cone-beam filtered backprojection reconstruction software. Improvements in image reconstruction methods, including use of CT attenuation maps and iterative reconstruction methods, should allow even more accurate in vivo quantitation of radioisotope uptake [1, 7, 9, 10, 33].
Conclusion Micro-SPECT/CT can be used to accurately quantify intratumoral radioisotope uptake in vivo and is more reliable than planar or micro-SPECT imaging alone, because of its ability to more accurately define tumor margins and regions of interest. Because combination SPECT/CT scanners are already being used in clinical practice, this method of quantitative image analysis is translatable and should be useful in the clinical setting.
332
S.K. Carlson et al.: Small Animal Micro-SPECT/CT Quantitation
Acknowledgments. This work was supported in part by the National Cancer Institute (grants K08 CA103859-01A1 and CA 100634-01), the Mayo Clinic SPORE in Pancreatic Cancer (grant P20 CA 102701), the Society of Gastrointestinal Radiology Research Grant Program, and the GE-AUR Radiology Research Academic Fellowship (GERRAF) Program. The authors thank Tracy Decklever for technical and imaging assistance, and doctoral student Dan Mundy for his valuable assistance with phantom experiments.
References 1. King MA, Pretorius PH, Farncombe T, Beekman FJ (2002) Introduction to the physics of molecular imaging with radioactive tracers in small animals. J Cell Biochem Suppl 39:221Y230 2. Beekman FJ, van der Have F, Vastenhouw B, et al. (2005) U-SPECT-I: a novel system for submillimeter-resolution tomography with radiolabeled molecules in mice. J Nucl Med 46(7):1194Y1200 3. Trotter DEG, Manjeshwar RM, Doss M, et al. (2004) Quantitation of small-animal 124I activity distributions using a clinical PET/SPECT scanner. J Nucl Med 45:1237Y1244 4. Pen˜uelas I, Haberkorn U, Yaghoubi S, Gambhir SS (2005) Gene therapy imaging in patients for oncological applications. Eur J Nucl Med Mol Imaging. (Available at: http://www.springerlink.com/media/ fckprmyxqhjmme2jqjft/contributions/n/8/0/7/n8076521873270j2_html/ fulltext.html. Accessed September 23, 2005) 5. Metzler SD, Jaszczak RJ, Patil NH, Vemulapalli S, Akabani G, Chin BB (2005) Molecular imaging of small animals with a triple-head SPECT system using pinhole collimation. IEEE Trans Med Imaging 24(7):853Y862 6. Yang D, Han L, Kundra V (2005) Exogenous gene expression in tumors: noninvasive quantification with functional and anatomic imaging in a mouse model. Radiology 235:950Y958 7. Tsui BMW (1996) The AAPM/RSNA physics tutorial for residents: physics of SPECT. Radiographics 16:173Y183 8. Ljungberg M, Sjo¨green K, Liu X, Frey E, Dewaraja Y, Strand S-E (2002) A 3-dimensional absorbed dose calculation method based on quantitative SPECT for radionuclide therapy: evaluation for 131I using Monte Carlo simulation. J Nucl Med 43:1101Y1109 9. King MA, Farncombe T (2003) An overview of attenuation and scatter correction of planar and SPECT data for dosimetry studies. Cancer Biother Radiopharm 18(2):181Y190 10. Meikle SR, Kench P, Kassiou M, Banati RB (2005) Small animal SPECT and its place in the matrix of molecular imaging technologies. Phys Med Biol 50:R45YR61 11. Hwang AB, Hasegawa BH (2005) Attenuation correction for small animal SPECT imaging using x-ray CT data. Med Phys 32(9): 2799Y2804 12. Hasegawa BH, Wong KH, Iwata K, et al. (2002) Dual-modality imaging of cancer with SPECT/CT. Technol Cancer Res Treat 1(6):449Y458 13. NCRP (1988) National Council on Radiation Protection and Measurements. Quality Assurance for Diagnostic Imaging Equipment. Publication 99. Bethesda, MD 14. Spitzweg C, Dietz AB, O_Connor MK, et al. (2001) In vivo sodium iodide symporter gene therapy of prostate cancer. Gene Ther 8:1524Y1531 15. Dingli D, Russell SJ, Morris JC (2003) In vivo imaging and tumor therapy with the sodium iodide symporter. J Cell Biochem 90(6):1079Y1086
16. DeNardo GL, Shen S, DeNardo SJ, et al.(1998) Quantification of iodine-131 in tumors using a threshold based on image contrast. Eur J Nucl Med 25:497Y502 17. Bland JM, Altman DG (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1(8476):307Y310 18. Lin LI (1989) A concordance correlation coefficient to evaluate reproducibility. Biometrics 45:255Y268 19. Koch W, Radau PE, Hamann C, Tatsch K (2005) Clinical testing of an optimized software solution for an automated, observer-independent evaluation of dopamine transporter SPECT studies. J Nucl Med 46:1109Y1118 20. Fleming JS, Bolt L, Stratford JS, Kemp PM (2004) The specific uptake size index for quantifying radiopharmaceutical uptake. Phys Med Biol 49:N227YN234 21. Scarfone C, Jaszczak RJ, Gilland DR, et al. (1999) Quantitative pulmonary single photon emission computed tomography for radiotherapy applications. Med Phys 26(8):1579Y1588 22. Dewaraja Y, Wilderman SJ, Ljungberg M, Koral KF, Zasadny K, Kaminiski MS (2005) Accurate dosimetry in 131I radionuclide therapy using patient-specific, 3-dimensional methods for SPECT reconstruction and absorbed dose calculation. J Nucl Med 46(5):840Y849 23. NRC (2004) U.S. Nuclear Regulatory Commission: Medical use of byproduct material. Code of Federal Regulations. 10 CFR 35. Washington, DC: U.S. Government Printing Office 24. Mattsson S, Jacobsson L, Johansson L (2003) Internal radionuclide dosimetry: diagnostic and therapeutic nuclear medicine, occupation and environmental exposure. Differences and similarities. Cancer Biother Radiopharm 18(3):423 25. Lambert B, Van de Wiele C (2005) Treatment of hepatocellular carcinoma by means of radiopharmaceuticals. Eur J Nucl Med Mol Imaging 32:980Y989 26. Imam S (2001) Status of radioimmunotherapy in the new millennium. Cancer Biother Radiopharm 16(3):244 27. Behr T, Behe M, Sgouros G (2002) Correlation of red marrow radiation dosimetry with myelotoxicity: empirical factors influencing the radiation-induced myelotoxicity of radiolabeled antibodies, fragments and peptides in pre-clinical and clinical settings. Cancer Biother Radiopharm 17(4):446 28. Knoop BO, Geworski L, Hofmann M, Munz DL, Knapp WH (2002) Use of recovery coefficients as a test of system linearity of response in positron emission tomography (PET). Phys Med Biol 47:1237Y1254 29. Ishimori T, Tatsumi M, Wahl RL (2005) Tumor response assessment is more robust with sequential CT scanning than external caliper measurements. Acad Radiol 12:776Y781 30. Iwata K, Vandehei T, Zhou H, et al. (2004) Radiotracer quantification with microSPECT-CT: phantom study [Abstract 128]. Mol Imaging 3(3):213 31. Zhang L, Iwata K, Vandehei T, et al. (2005) Radiotracer quantification with in vivo preclinical microSPECT/CT [abstract 329]. Mol Imaging Biol 7(2):166 32. Tatsumi M, Nakamoto Y, Traughber B, Marshall LT, Geschwind J-FH, Wahl RL (2003) Initial experience in small animal tumor imaging with a clinical positron emission tomography/computed tomography scanner using 2-[F-18]Fluoro-2-deoxy- D -glucose. Cancer Res 63:6252Y6257 33. Koch W, Hamann C, Welsch J, Po¨pperl G, Radau PE, Tatsch K (2005) Is iterative reconstruction an alternative to filtered backprojection in routine processing of dopamine transporter SPECT studies? J Nucl Med 46(11):1804Y1811