Evaluation of simultaneous 201Tl/99mTc dual-isotope cardiac SPECT imaging with model-based crosstalk compensation using canine studies

Simultaneous Tl-201/Tc-99m-sestamibi dual-isotope myocardial perfusion SPECT imaging can reduce imaging time and produce perfectly registered rest/stress images. However, crosstalk from Tc-99m into Tl-201 images can significantly reduce Tl-201 image quality. We have developed a model-based compensation (MBC) method to compensate for this crosstalk. The method has previously been validated with phantom and simulation studies. In this study, we evaluated the MBC method using a canine model. Left anterior descending or left circumflex coronary artery stenoses were created in 50 adult mongrel dogs weighing 20-30 kg. The dogs were injected with 111 MBq (3 mCi) of Tl-201 at rest, and a SPECT study acquired. Stress was induced by administering adenosine to the dog, followed by injection of 740 MBq (20 mCi) of Tc-99m-sestamibi at peak stress. A second SPECT study was performed with data acquired in both Tl-201 and Tc-99m energy windows to provide simultaneous dual-isotope projection data. The images were reconstructed using the ordered-subsets expectation-maximization reconstruction algorithm with compensation for attenuation, scatter, and detector response. For simultaneously acquired Tl-201 data, we also applied the MBC method to compensate for crosstalk contamination from Tc-99m. Without compensation, Tc-99m crosstalk increased the estimated Tl-201 activity concentration in the rest images and reduced defect contrast. After MBC, the Tl-201 images were in good agreement with the registered single-isotope images and ex vivo count data. The ischemic (IS) to non-ischemic (NIS) region Tl-201 activity concentration ratios were computed for single-isotope and dual-isotope studies. The correlation with ex vivo IS-NIS ratios was 0.815 after MBC, compared to the 0.495 from data without compensation. In addition, the regression line for the IS-NIS ratios with MBC was almost parallel to the line of identity with a slope of 0.93, compared to a slope of 0.45 without compensation. These results demonstrate that model-based crosstalk compensation can provide substantial reduction of crosstalk effects in simultaneously acquired myocardial perfusion SPECT images in living biological systems.