Accuracy of absolute activities recovered by three commercial SPECT-CT reconstruction packages using different quantitation techniques

Objectives Advancements in SPECT-CT technology raise important questions regarding the quality and accuracy of reconstructed images. This multi-center phantom study compares the activities obtained from three modern SPECT-CT systems with their most advanced software packages using three quantitation approaches.

Methods A thorax phantom with a normal 120mL heart and two 33mL bottles filled with Tc-99m solution (clinically relevant concentrations) was scanned (reconstructed) using: Siemens Symbia-T6 (Flash3D^TM), GE Infinia Hawkeye (Evolution^TM), and Philips Precedence (Astonish^TM). Images were rescaled to absolute activities. Two simple segmentation methods were used to estimate activities by summing the counts within an object’s boundaries determined by: (A) a threshold set to the level of the background and (B) a CT-based exact volume. The third method (C) quantifies activity by recovering spill-out counts and adding them to the counts within the exact volume.

Results Resolution of the reconstructed images differed substantially. Quantitatively, method A overestimated the volumes (up to 410%) and the activities (up to 135%), while the results for method B varied largely between the tested software packages. Activities obtained using method C remained within 10% of the true values with the Astonish reconstruction producing the most accurate results.

Conclusions Reconstructed images differed and quantitation based on CT did not provide reliable results due to its dependence on image resolution, thus more advanced methodologies (such as method C) must be used.