Quantitative accuracy of sequential dual-isotope SPECT/CT reconstructions

Objectives We examine an ability of sequential reconstruction algorithms to recover absolute tracer activity from dual-isotope SPECT/CT studies with different ratios of Tc-99m/In-111 (Exp.1) and Tc-99m/I-123 concentrations (Exp.2).

Methods In our first series of experiments, four 30ml containers with different ratios of Tc-99m and In-111 activities were placed inside the Jaszczak phantom which was (i) kept empty and (ii) filled with cold water and scanned using a hybrid SPECT/CT system (Infinia Hawkeye, GE Healthcare). In the second series, the same acquisitions with Tc-99m and I-123 tracers were made. The data was collected in 20% energy windows set on all photopeaks. Our sequential reconstruction algorithm iteratively performs the following steps: (1) reconstructs two separate images from each isotope’s own energy windows; (2) models all scatter and cross-talk components using these images and Klein-Nishina formula; (3) incorporates these corrections into the forward step of OSEM and updates images. The absolute activity of each tracer was measured in regions of interest encompassing each container.

Results The reconstructed absolute activities of In-111 images (Exp.1) were within 0.8-7.0% from the truth; activities of I-123 images (Exp.2) had errors 0.1-8.3%; and Tc-99m activities (Exp.1 and Exp.2) - 0.1-13.1%. The number of required iterations depended on the severity of cross-contamination. While only single modeling of down-scatter was required in Exp.1 (large energy difference between isotopes), two updates were necessary for mutually contaminated Tc-99m/I-123 data (Exp.2).

Conclusions Accurate model-based corrections for scatter and cross-talk with CT attenuation maps allowed us to recover absolute activities from SPECT/CT scans with errors comparable to those in standard single-isotope studies. The proximity of energies and unfavorable activity ratios increased the computation time, but practically did not affect the resulting accuracy. Critical parameters influencing performance of our method are qualities of attenuation map and of down-scatter model