Evaluation of how the number of delineated vertebrae in SPECT influences the accuracy of bone marrow dosimetry

Introduction: Peptide receptor radionuclide therapy using ¹⁷⁷Lu-DOTATATE is an established treatment option for somatostatin-receptor positive metastatic neuroendocrine tumors (NETs). Although the main dose-limiting organ in ¹⁷⁷Lu-DOTATATE treatment are the kidneys, bone marrow toxicity is more common than renal toxicities. SPECT-based activity quantification in small bone marrow cavities is challenging due to high noise levels combined with low signal. In addition, the absorbed dose can be overestimated by spill-in counts from activity in organs nearby. A way to reduce the spill-in counts is to increase the number of updates in the SPECT reconstruction. However, increasing the number of updates will also increase the noise level, potentially reducing the accuracy in bone marrow dosimetry. One way of improving the accuracy is to use multiple bone marrow sites in the dosimetric calculations.

The aim of this study was to examine how the number of updates in the SPECT reconstruction and the number of vertebrae will influence the accuracy in bone marrow dosimetry.

Methods: This is a retrospective analysis of 13 consecutive patients diagnosed with NETs treated with ¹⁷⁷Lu-DOTATATE (7400 MBq x 4 cycles). Following the first cycle, SPECT/CT imaging was performed at 2, 24, 48 and 168 h post injection using 120 projections. These four SPECT scans were reconstructed using the Monte Carlo-based OSEM reconstruction code SARec, with 1-200 updates where every 10^th update was saved. Eight bone marrow cavities (T10–L5) were segmented, and the coefficient of variation (COV) was calculated for each time point. The COV was calculated using 1, 3, 5 and 7 vertebrae, to evaluate how the number of vertebrae influenced quantification of absorbed dose.

Results: For most vertebrae, the concentration in the cavity decreased during the first ~50 updates but then stabilizes, see Figure 1. Thereafter, the COV increased with increased number of updates. At 60 updates, the median and range of COV for all patients activity concentration at timepoints 2, 24, 48 and 168 h were, respectively, 0.25 (0.14–0.52), 0.31 (0.16–0.60), 0.35 (0.25–0.63) and 0.35 (0.17–0.58). The time integrated activity median and range of COV, using 1, 3, 5 and 7 vertebrae were 0.30 (0.21–0.48), 0.16 (0.07–0.22), 0.06 (0.03–0.11) and 0.04 (0.03–0.07), respectively. The median and range of the bone marrow dose was estimated to 0.06 (0.03–0.11) Gy/GBq.

Conclusions: When reconstructing SPECT data, the spill-in counts from organ activities is reduced up to roughly 60–100 updates for the vertebrae. An optimal number of updates, which balanced the spill-in influence and the increasing COV with increased number of updates, was around 60 for the used reconstruction protocol. The results also demonstrated that multiple bone marrow cavities are required for obtaining acceptable dosimetry accuracy.

• Download figure
• Open in new tab
• Download powerpoint