Cross-talk correction for dual-isotope imaging with a dedicated cardiac SPECT camera

Simultaneous dual-isotope perfusion imaging has the potential to reduce protocol durations and improve image alignment but requires cross-talk correction. The Discovery NM 530c dedicated cardiac camera offers two possible advantages over traditional cameras with respect to cross-talk. The cadmium zinc telluride (CZT) detectors have improved energy resolution and the pinhole collimators lead to reduced lead-X-ray production. The objective of this study is to determine the crosstalk fractions and blurring kernels required for ^99mTc/²⁰¹Tl rest/stress imaging. Forty-six pairs of single-isotope clinical ^99mTc-tetrofosmin and ²⁰¹Tl studies were examined. The patients were matched for height, weight, and gender. Listmode data (8min) were reprocessed to obtain the signal in the following windows: Tc-10 (140keV±10%), Tc-8 (140keV±8%), Tc-6 (140keV±6%), Sc (100keV±10%), Tl-167 (167keV±8%), Tl-70a (70keV±15%), and Tl-70b (70keV-15%+10%). The relative counts in each window were determined for both isotopes for rest and stress imaging. Symmetric 2D Gaussian kernels were found that minimized the squared difference between convolved photon distributions from one window and corresponding measured cross-talk. 14 sets of simulated dual-isotopes studies were created and the correction assessed by comparing the summed stress scores before and after correction. ^99mTc contributes ≤1% of the counts in the Tl-167 window and 40-80% of the counts in the Tl-70 window. ²⁰¹Tl contributes 1-10% of the counts in the ^99mTc-photopeak windows. Crosstalk contamination is less with the narrower energy windows that capitalize on the improved energy resolution of the CZT system. The weighting factors for Tc-6 and Sc windows for optimal estimation of the Tl-70b window interference were: -0.24:1.24 with Gaussian σ=0.58 and 1.2 pixels respectively. Considering the 14 - - sets of simulated dual-isotope studies, the summed stress scores for the corrected data correlated more closely with truth than did the uncorrected data. The multi-energy window convolution subtraction approach has promise for correcting cross-talk in ²⁰¹Tl/^99mTc simultaneous dual-isotope myocardial perfusion imaging with a dedicated cardiac SPECT system.