Preoperative parathyroid scintigraphic lesion localization: accuracy of various types of readings.

PURPOSE: To retrospectively compare the accuracy of various parathyroid scintigraphy readings for single-gland disease (SGD) and multigland disease (MGD) in patients with primary hyperparathyroidism, with histologic analysis as the reference standard. MATERIALS AND METHODS: Institutional review board approval was obtained for this HIPAA-compliant study. Records of 462 patients with primary hyperparathyroidism who underwent preoperative imaging with a technetium 99m ((99m)Tc) sestamibi and (99m)TcO4- protocol that consisted of early and late pinhole (99m)Tc sestamibi, pinhole thyroid imaging, image subtraction, and single photon emission computed tomography (SPECT) were retrospectively reviewed. An experienced nuclear medicine physician without knowledge of other test results or of the final diagnoses graded images on a scale from 0 (definitely normal) to 4 (definitely abnormal). Early pinhole (99m)Tc sestamibi images, late pinhole (99m)Tc sestamibi images, subtraction images, SPECT images, early and late pinhole (99m)Tc sestamibi images, all planar images, and all images--including SPECT images--were read in seven sessions. Receiver operating characteristic curves were generated for each session and were used to calculate sensitivity, specificity, and accuracy. RESULTS: A total of 534 parathyroid lesions were excised. Of the 462 patients, 409 had one lesion, whereas 53 had multiple lesions. Reading all images together was more accurate (89%, P = .001) than was reading early (79%), late (85%), subtraction (86%), and SPECT (83%) images separately; however, it was not significantly more accurate than reading planar images (88%) or early and late images together (87%). Reading all images was significantly less sensitive in the detection of lesions with a median weight of 600 mg or less than in the detection of lesions with a median weight of more than 600 mg (86% vs 94%, P = .004). Per-lesion sensitivity for reading all images was significantly higher for SGD than for MGD (90% vs 66%, P < .001). Sensitivity of reading all images together in the identification of patients with MGD was 62%. CONCLUSION: Reviewing early, late, and subtraction pinhole images together with SPECT images maximizes parathyroid lesion detection accuracy. Test sensitivity is adversely affected by decreasing lesion weight and MGD.

Phantom experiments to improve parathyroid lesion detection.

This investigation tested the hypothesis that visual analysis of iteratively reconstructed tomograms by ordered subset expectation maximization (OSEM) provides the highest accuracy for localizing parathyroid lesions using 99mTc-sestamibi SPECT data. From an Institutional Review Board approved retrospective review of 531 patients evaluated for parathyroid localization, image characteristics were determined for 85 99mTc-sestamibi SPECT studies originally read as equivocal (EQ). Seventy-two plexiglas phantoms using cylindrical simulated lesions were acquired for a clinically realistic range of counts (mean simulated lesion counts of 75 +/- 50 counts/pixel) and target-to-background (T:B) ratios (range = 2.0 to 8.0) to determine an optimal filter for OSEM. Two experienced nuclear physicians graded simulated lesions, blinded to whether chambers contained radioactivity or plain water, and two observers used the same scale to read all phantom and clinical SPECT studies, blinded to pathology findings and clinical information. For phantom data and all clinical data, T : B analyses were not statistically different for OSEM versus FB, but visual readings were significantly more accurate than T : B (88 +/- 6% versus 68 +/- 6%, p = 0.001) for OSEM processing, and OSEM was significantly more accurate than FB for visual readings (88 +/- 6% versus 58 +/- 6%, p < 0.0001). These data suggest that visual analysis of iteratively reconstructed MIBI tomograms should be incorporated into imaging protocols performed to localize parathyroid lesions.