A commissioning workflow for quantitative ¹⁷⁷Lu SPECT/CT across different Siemens scanner–reconstruction configurations

Abstract

The accuracy of quantitative lutetium-177 single-photon emission computed tomography/computed tomography (¹⁷⁷Lu SPECT/CT) depends on the imaging system, acquisition protocol, reconstruction algorithm, and calibration strategy. This study aimed to develop and evaluate a practical commissioning workflow to optimize ¹⁷⁷Lu imaging across different commercial scanner–reconstruction configurations. Three Siemens workflows were evaluated: FLASH3D and xSPECT Quant on the Symbia Intevo Bold, and FLASH3D+ on the Symbia Pro.specta. A uniform phantom was used to assess image noise and derive calibration factors for the ordered-subset expectation maximization (OSEM)-based reconstructions. Recovery coefficients (RCs) were evaluated via an International Electrotechnical Commission/National Electrical Manufacturers Association phantom at various sphere-to-background ratios and modeled using a three-parameter logistic function for partial volume correction. Reconstruction parameters were optimized by jointly evaluating RC convergence and image noise, followed by quantitative validation using an anthropomorphic phantom. OSEM-based workflows showed a predictable dependence on equivalent iterations, yielding lower image noise. Conversely, xSPECT Quant exhibited complex parameter dependence and higher noise, but provided superior activity recovery—with RCs closer to unity—and the lowest quantification errors during validation. A reconstruction setting of two subsets and 30 iterations was selected for all workflows. For FLASH3D and FLASH3D+, this configuration provided RCs closest to unity while maintaining the coefficient of variation below the predefined 15% threshold. The same setting was adopted for xSPECT Quant to ensure methodological consistency despite its intrinsically higher image noise. This commissioning workflow provides a practical framework for local ¹⁷⁷Lu SPECT/CT optimization. Among the configurations, xSPECT Quant yielded higher recovery coefficients and lower quantification errors, albeit with increased noise. These findings highlight the need for locally optimized protocols for patient-specific dosimetry and do not establish the general superiority of any one reconstruction algorithm.