Gating provides improved accuracy for differentiating artifacts from true lesions in equivocal fixed defects on technetium 99m tetrofosmin perfusion SPECT.

Journal Article (Journal Article)

BACKGROUND: Because equivocal fixed perfusion defects on single photon emission computed tomography (SPECT) often cause a diagnostic dilemma in patients with suspected coronary artery disease (CAD), we used receiver-operating characteristic (ROC) curve analysis to evaluate the degree of increased accuracy provided by adding gated images for interpretation of such cases. METHODS AND RESULTS: One hundred five (29%) of 365 consecutive patients undergoing technetium 99m tetrofosmin gated SPECT for evaluation of CAD demonstrated an equivocal fixed defect. Two independent observers graded the probability for true lesion with a 5-point system over 3 steps of interpretation: reviewing tomographic images alone (step 1), adding projection images (step 2), and adding gated cine images (step 3). Based on clinical criteria, 66 patients were categorized as disease negative and 25 as disease positive, while the final diagnosis was undetermined in 14 cases. Diagnostic performance was significantly higher at step 3 than at step 1 for both observers (p < 0.05 and < 0.0001) and at step 2 for observer B (p < 0.005). The maximum accuracy increased from 78.0% to 80.2% at step 2 to 84.6% to 86.8% at step 3. Moreover, the agreement of interpretation between the 2 observers was higher at step 3 (kappa = 0.53) than at step 1 (kappa = 0.29) or step 2 (kappa = 0.25). CONCLUSION: In patients showing an equivocal fixed defect on 99mTc tetrofosmin SPECT, gated cine images provide improved differential accuracy and enhanced objectivity of interpretation by reducing interobserver variance.

Full Text

Duke Authors

Cited Authors

  • Choi, JY; Lee, KH; Kim, SJ; Kim, SE; Kim, BT; Lee, SH; Lee, WR

Published Date

  • January 1, 1998

Published In

Volume / Issue

  • 5 / 4

Start / End Page

  • 395 - 401

PubMed ID

  • 9715984

International Standard Serial Number (ISSN)

  • 1071-3581

Digital Object Identifier (DOI)

  • 10.1016/s1071-3581(98)90145-3


  • eng

Conference Location

  • United States