Theranostic SPECT Reconstruction for Improved Lesion Dosimetry in Radionuclide Therapy

Introduction: We have exploited the theranostic approach of diagnostic PET imaging followed by SPECT imaging of the therapeutic radionuclide to improve the resolution and noise characteristics of the SPECT reconstruction to improve dosimetry estimations. This novel approach to SPECT reconstruction uses the PET image, with its superior resolution, to guide the SPECT reconstruction. We term this reconstruction approach SPECTRE (Single Photon Emission Computed Theranostic REconstruction).

Methods: The SPECTRE reconstruction incorporating PET images as side-information uses the Hybrid Kernelized Expectation-Maximisation (HKEM) algorithm implemented in STIR, a software library for tomographic image reconstruction. We first demonstrate SPECTRE in a dual PET-SPECT IEC phantom study using ⁶⁸Ga/¹⁷⁷Lu where Recovery Coefficients (RC) for each of the reconstructions was calculated in order to assess quantitative accuracy. We then applied this to a patient theranostic study using ⁶⁴Cu/⁶⁷Cu SARTATE in multifocal meningioma and compared it with conventional SPECT image reconstruction methods. The patient received 177 MBq of [⁶⁴Cu]SARTATE for the PET study and was scanned at 1, 4 and 24 hrs post-injection. The SPECT images used were from Cycle 1 of therapy with 5.1 GBq of [⁶⁷Cu]SARTATE and the 4 hr post-injection time point was used for the comparison. The corresponding 4 hr ⁶⁴Cu PET image was used as the side-information in the SPECTRE reconstruction. Other quantitative SPECT reconstructions used conventional OSEM with CT-based scatter and attenuation correction (“qSPECT”) without resolution modelling and the STIR OSEM implementation with Resolution Modelling (“OSEM+RM”). Volumes of Interest (VOIs) were defined on the PET image and propagated to the SPECT reconstructions where SUV_mean and SUV_max were calculated for direct comparison.

Results: Figure 1 A) shows the same transverse slice for each reconstruction method and B) shows a plot of the mean Recovery Coefficient in each sphere. OSEM+RM demonstrates improved concentration recovery over qSPECT at the price of increasing noise. SPECTRE on the other hand shows a similar improvement in recovery but with better noise characteristics. Figure 1 C) shows a Maximum Intensity Projection (MIP) of the 4hr ⁶⁴Cu PET reconstruction with each VOI labelled (top left). Next to this, the same transverse slices for qSPECT and SPECTRE reconstructions are displayed; showing lesions 1, 3 and 4. The same windowing has been applied to both SPECT reconstructions. Compared to the PET image, the qSPECT reconstruction underestimates the activity in the lesions, while the SPECTRE reconstruction has comparable recovery to the PET.

Conclusions: Methods that incorporate RM into the reconstruction quantitatively perform much better than standard reconstruction methods that do not attempt to mitigate the poor spatial resolution. In the clinical example the SPECTRE approach saw an increase in SUV_mean in 5 target lesions by a factor of 2-3 over the conventional qSPECT reconstruction. The SPECTRE and OSEM+RM reconstructions both had SUV_mean and SUV_max values comparable to those in the PET image, however, the SPECTRE reconstruction had far superior noise properties without Gibbs artefacts. Acknowledgements: We are grateful to Clarity Pharmaceuticals (Sydney, Australia), in particular Michelle Parker and Dr Mathew Harris, for their support and access to the ⁶⁴Cu/⁶⁷Cu SARTATE image data (ClinicalTrials.gov Identifier: NCT03936426). The authors would like to thank the STIR consortium for their ongoing support. Figure 1. A) IEC (Left top) ⁶⁸Ga PET, TOF+RM+5mm Gaussian. (Middle top) qSPECT (4it, 8s). (Left bottom) OSEM+RM (40it, 12s). (Middle bottom) SPECTRE+RM (40it, 12s). B) RC_mean in phantom spheres. C) Clinical example (Top left) MIP of 4hr ⁶⁴Cu-PET head with VOIs labelled. (Top right) ⁶⁴Cu-PET: TOF+RM+5 mm Gaussian Filter. ⁶⁷Cu-SPECT: (Bottom Left) qSPECT (4it, 8s), (Bottom right) SPECTRE+RM (40it, 12s). L - Lesion; P - Parotids

• Download figure
• Open in new tab
• Download powerpoint