Characterization of noise and resolution for quantitative 177 Lu SPECT/CT with XSPECT quant

Abstract

Quantitative SPECT/CT imaging forms the basis for internal dosimetry in molecular radiotherapies. While the conversion from counts to activity is typically based on conversion factors individually measured by each site, a recently introduced commercially available reconstruction (xSPECT Quant) offers a standardized and traceable calibration of SPECT/CT systems. The aim of this work was to assess the characteristics of xSPECT Quant in combination with 177 Lu as one of the most important radionuclides used in molecular radiotherapies and to compare it to a widely used ordered-subset expectation-maximization reconstruction (Flash3D). Methods: In a series of 177 Lu-filled phantom measurements, several important features were investigated for xSPECT Quant and Flash3D. Noise behavior and accuracy of the activity determination were evaluated in a large cylinder. Recovery coefficients were assessed in a hot-sphere phantom with and without background. Additionally, the resolutions were determined in a line source phantom as well as in a matched-filter resolution analysis of the hot-sphere-cold-background phantom. Results: Both reconstruction algorithms improve the spatial resolution at the cost of noise buildup. Despite its slower convergence, Flash3D features a more efficient recovery. Although resolution recovery methods are applied within both reconstructions, partial-volume errors-namely activity overestimation in the object center and spill-out of counts from the object edges-remain of relevance. In contrast to Flash3D, for which only the total number of updates (iterations · subsets) plays a role, the exact subdivision into iterations and subsets affected all characteristics of xSPECT Quant (optimum, 1 subset). The optimal trade-off between noise build-up and resolution improvement was found for 48 iterations and 1 subset, resulting in a quantitative accuracy of 1.2% in the Jaszczak cylinder (xSPECT Quant cross-calibrated to the dose calibrator). Conclusion: If the reconstruction parameters are chosen with care, both examined reconstructions can provide absolute quantitative SPECT images with high image quality (subcentimeter resolution at an acceptable noise build-up) as well as high quantitative accuracy (given a well-calibrated Flash3D conversion from counts to activity concentration). With its standardized (and traceable) activity determination, xSPECT Quant dispenses with site-specific calibration protocols, enabling a reliable activity determination comparable across sites, which is especially useful for multicentric molecular radiotherapy studies.