Preoperative ¹²³I/^99mTc-Sestamibi Subtraction SPECT and SPECT/CT in Primary Hyperparathyroidism

Patients

The study group consisted of 18 men and 43 women with a mean age of 59 y (range, 18–85 y). Before surgery, the median total serum calcium was 11.0 mg/dL (range, 8.8–14.5 mg/dL; reference level, 8.5–10.5 mg/dL), and the median serum PTH was 114 pg/mL (range, 62–768 pg/mL; reference level, 10–60 pg/mL).

Imaging

Four hours after the oral administration of a mean dose of ¹²³I sodium iodide (12,021.3 kBq; range, 10,915–12,765 kBq [324.9 μCi; range, 295–345 μCi]), each patient received a mean dose of ^99mTc-sestamibi intravenously (1,191.4 MBq; range, 1,106.3–1,224.7 MBq [32.2 mCi; range, 29.9–33.1 mCi]). Ten minutes after the sestamibi administration, imaging began. SPECT/CT data were acquired with a Symbia T6 scanner (Siemens Molecular Imaging). The SPECT emission data were collected in dual-energy windows in an attempt to separate the ^99mTc and ¹²³I counts. The ^99mTc window was centered at 140 keV and had 15% width (range, 129.5–150.5 keV). The ¹²³I window was placed with a 4% offset above 159 keV and had 15% width (range, 153.4–177.3 keV). The 4% offset was selected to minimize the spillover of the ^99mTc photopeak into the much less intense ¹²³I photopeak. Before imaging, ^99mTc and off-peak ¹²³I flood images were obtained with the SPECT camera, and all these flood images appeared visually uniform. Emission data were acquired in a step-and-shoot sequence with a contoured noncircular orbit (dual detectors in 180° configuration; low-energy high-resolution collimators; matrix, 128 × 128; pixel size, 4.8 mm; 60 steps per each detector; 3°/step; 25 s/step; total scan time, 25 min). After the SPECT emission scan, the patient remained on the table for the CT portion of the examination; no oral or intravenous contrast agents were administered. After a topogram scout scan (130 kVp, 100 mA, anterior view), a helical CT scan was performed (130 kVp, 6 × 2.0 mm collimation, 0.8-s rotation time, 1.0 pitch). The dose was reduced by tube-current modulation (CARE Dose; Siemens), with the reference exposure set to 80 mAs. Two sets of images were reconstructed. The first set of images, for diagnostic purposes, was reconstructed with a slice thickness of 5 mm at 2.5-mm increments, with a B31s (medium smooth +) filter kernel. The second set of images, used for SPECT attenuation correction, was reconstructed with a slice thickness of 5 mm at 5-mm increments, with a B08s filter kernel. SPECT images were reconstructed on the scanner console workstation using the iterative ordered-subsets algorithm including depth-dependent collimator blur modeling. First, the raw projection data were inspected for patient motion during the scan; if motion was present, the projections were manually shifted to minimize the motion. Next, an initial (non–attenuation-corrected) SPECT reconstruction was performed for each of the 2 acquisition windows. The image-reconstruction settings (FLASH3D software; Easypano) were 4 iterations, 8 subsets, and 8 mm.

Gaussian Postprocessing Filter

The non–attenuation-corrected SPECT and fused SPECT/CT images were displayed to check the accuracy of coregistration. If patient motion was seen between the SPECT and CT images, the CT images were manually shifted to minimize the misregistration. Afterward, the CT images were downsampled to match the SPECT matrix and converted from Hounsfield units into effective attenuation values at 140 keV (^99mTc) and 159 keV (¹²³I). Final SPECT images were then reconstructed with attenuation correction using these CT-derived attenuation maps. The image-reconstruction settings (FLASH3D software) were 4 iterations, 8 subsets, and an 8-mm gaussian postprocessing filter.

Analysis

All studies in this series were interpreted visually in a prospective manner, before surgery, by an experienced nuclear radiologist unaware of the results of any other localization studies. The non–attenuation-corrected SPECT images were evaluated first. ¹²³I SPECT data were first normalized to thyroid tissue on the ^99mTc SPECT data using the image arithmetic tools available on the clinical workstation. An initial normalization factor was determined as the inverse ratio of the maximum voxel value corresponding to thyroid tissue on the ¹²³I SPECT to the value of the corresponding voxel in the ^99mTc SPECT dataset. Each voxel in the ¹²³I SPECT dataset was multiplied by this normalization factor. The normalized ¹²³I SPECT data were then subtracted from the ^99mTc SPECT data to create a subtraction SPECT dataset. In an iterative fashion, the operator was able to adjust this normalization factor until, by visual inspection, subjectively satisfactory subtraction SPECT images were obtained, which were used for interpretation (Fig. 1). A focal site of residual activity on the subtraction SPECT images was defined as a positive finding. Each positive finding was assigned an anatomic location (right or left neck, upper or lower thyroid bed, anterior or posterior superior mediastinum, or other anatomic site) and graded on a 5-point scale of diagnostic confidence (1, definitely negative; 2, probably negative; 3, equivocal; 4, probably positive; and 5, definitely positive for a parathyroid lesion).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

Selected series of consecutive coronal dual-isotope SPECT tomograms from scans performed on 61-y-old man with primary hyperparathyroidism. Subtraction of normalized ¹²³I window tomograms (B) from corresponding ^99mTc window tomograms (A) results in subtraction SPECT images (C), which demonstrate small residual focus of activity (arrows) in left lower neck. At surgery, residual focus of activity was found to represent 325-mg parathyroid adenoma.

At a separate reading session, attenuation-corrected SPECT/CT images were interpreted by the same observer. After normalization and subtraction of the attenuation-corrected ¹²³I SPECT data from the attenuation-corrected ^99mTc SPECT data, an attenuation-corrected subtraction SPECT dataset was created, which was used for interpretation. Coregistration of the attenuation-corrected subtraction SPECT images with the CT images was performed using fusion software. The operator was able to adjust the window settings, color bars, and mixing ratios of the fused images. A positive finding was defined as focal residual activity on the attenuation-corrected subtraction SPECT image that corresponded to a nonthyroidal soft-tissue lesion on CT images judged to be in an anatomic location consistent with a parathyroid lesion. Each positive finding was assigned this anatomic location and graded on the same 5-point scale as was used for the non–attenuation-corrected SPECT images.

Parathyroid surgery was performed on each patient by experienced endocrine surgeons who were aware of the subtraction SPECT and the SPECT/CT results. The appearance, location, and size of each suspected parathyroid lesion was documented. Histopathologic examination was performed, and weights of all resected surgical specimens were obtained. To estimate sensitivity and specificity for each of the 2 imaging techniques, scores of 1–3 were considered negative and scores of 4–5 were considered positive. The McNemar test, adjusted for the clustered data (i.e., multiple glands per patient) (13), was used to compare the sensitivities of subtraction SPECT and SPECT/CT and also to compare their specificities. A significance level of 0.05 was used. Receiver-operating-characteristic (ROC) curves were also constructed for each of the 2 modalities. The areas under the ROC curves of the 2 were compared using methods for clustered paired ROC data (14). Generalized linear models, using generalized estimating equations to handle the clustered data (i.e., multiple glands were evaluated in some patients), were fit to assess the association between lesion detectability and gland weight, patient calcium level, and patient PTH level. For each modality, the dependent variable in the model was whether the lesion was a true-positive (TP) or false-negative (FN); the independent variable in the model was lesion weight, calcium level, or PTH level. Because of the small sample size, a multiple-variable model could not be assessed.