Validation of the Yale circumferential quantification method using 201Tl and 99mTc: a phantom study.

Abstract

UNLABELLED: The Yale circumferential quantification (Yale CQ) method for quantification of SPECT images has been validated previously using empirically derived correction factors. In the present studies, the Yale CQ method was further validated using 2 SPECT gamma cameras and 2 radioisotopes. METHODS: SPECT images were acquired from cardiac phantoms with multiple fillable inserts to simulate myocardial perfusion defects of varying extents and severities. Seventy phantom configurations were created. One hundred and forty SPECT images (70 with 99mTc and 70 with 201TI) were acquired using a triple-head SPECT camera. SPECT defects were quantified using the Yale CQ method, with incorporation of 99mTc- and 201TI-derived normal databases and correction factors. RESULTS: Quantified phantom SPECT defect sizes acquired with 99mTc correlated well with actual calculated defect sizes (r = 0.96, y = 0.92x - 0.41). Bland-Altman analysis of agreement revealed strong agreement over a wide range of defect sizes, with a mean error of 1.2% and 2 SDs of 5.0%. Overall 201TI SPECT defect sizes also correlated well with actual defect sizes (r = 0.92), but there was a systematic underestimation (y = 0.72x - 0.76). Bland-Altman analysis showed underestimation over the entire range of defect sizes, with a mean error of 3.4% and 2 SDs of 7.5%. Implementation of a normal 201TI phantom database improved accuracy of quantification (r = 0.95, y = 0.87x - 1.36). The addition of 201TI-specific correction factors further improved accuracy (r= 0.94, y = 0.98x - 1.52). Reproducibility of SPECT defect sizes quantification for 99mTc using 2 gamma cameras was excellent (r = 0.98, y = 0.98x + 0.84). CONCLUSION: The Yale CQ SPECT quantification method, using the empirically derived correction factors, provides accurate and reproducible quantification of phantom defects over a wide range of defect sizes. Accurate quantification of 201TI and 99mTc SPECT defect sizes requires radiotracer-specific normal databases.