A projection-domain quantification method for absolute quantification with low-count SPECT for alpha-particle radiopharmaceutical therapy

Objectives: Quantitative SPECT provides a mechanism to measure radiation-absorbed doses in lesions and normal organs post administration of alpha-particle radiopharmaceutical therapies (α-RPTs) [1]. However, reliable dose quantification requires reliable absolute quantification of activity uptakes. This is especially challenging with α-RPTs due to the unique isotope physics and the very low number of detected gamma-ray photons (up to 20-fold less than conventional SPECT). At these low count levels, the conventional reconstruction-based quantification methods are highly erroneous [2-4]. To address these challenges, we developed a projection-domain quantification (PDQ) method for α-RPT SPECT. Methods: The proposed PDQ method directly estimates the regional activity uptake from the projection data, thus avoiding the information loss in reconstruction and also reducing the ill-posedness of the inverse problem by estimating a smaller number of regional activity uptakes as opposed to a large number of voxel values. The method extends a previously proposed maximum-likelihood region-of-interest method [5], making multiple advances to account for the unique challenges of quantification with α-RPTs. These include accurately modeling the isotope and imaging physics, including the effects of attenuation, scatter, and collimator-detector response using a Monte Carlo-based procedure, accounting for noise due to stray radiation that becomes relevant at these low counts, and developing the method in 3D. The method requires access to the definition of the various regions, as can be obtained from the patient CT, MRI, or PET scans. The method was validated in the context of 3D SPECT with Radium-223, a common radionuclide for α-RPTs [6]. Validation was performed using both clinically realistic simulation and physical-phantom studies. Before conducting simulations, their accuracy was validated. Next, a virtual imaging trial (VIT) was conducted simulating a population of patients with prostate cancer that had metastasized to the pelvic bone and who were being administered Ra-223 therapy. For this purpose, 50 digital 3D male patients (Subfig. 1 a - b) with different anatomies were generated using the XCAT phantom [7]. Projection data for these patients were obtained by simulating a GE Optima 640 SPECT system with a HEGP collimator using SIMIND [8]. Regional activity uptakes were estimated from the projection data using the proposed method. Next, we studied the sensitivity of the method to lesion size and lesion-to-bone uptake ratio (LBUR). In the physical-phantom study, a NEMA phantom (Subfig. 1c) was filled with clinically relevant Ra-223 activity concentrations and scanned on a GE Optima 640 SPECT/CT. Again, the regional activity uptakes were estimated from the projection data with the PDQ method and estimation accuracy was evaluated. Comparative tests with a conventional ordered subset expectation maximization (OSEM)-based reconstruction method was also performed.

Results: The proposed method yielded reliable estimates of regional uptakes and significantly outperformed the OSEM-based method in both realistic simulations and physical-phantom studies (Subfig. 2 and 3). In the VIT, the method yielded an average root mean square error (RMSE) of 5% over all regions (Subfig. 2a). Also, the method yielded reliable quantitation across lesion sizes and LBURs (~5% RMSE) (Subfig. 2 b - c). Similarly, in the NEMA-phantom study, the proposed method yielded an averaged 57% lower bias compared to OSEM-based method over the six spheres with diameters between 10 mm - 37 mm.

Conclusions: The proposed PDQmethod provided reliable absolute quantification of regional uptakes in SPECT at ultra-low count levels as validated using simulation and physical-phantom studies. The method could thus provide a mechanism for dose quantification with α-RPTs.