Lung Dose Verification Following Y90-Radioembolization Therapy: An Inter-modality Comparison

Purpose: Post-radioembolization 3D liver dosimetry is increasingly used to verify treatment delivery and predict treatment outcomes. However, lung shunt fraction (LSF) and dose estimates are primarily based on pre-therapy 2D imaging of surrogate Tc99m-MAA particles. The objective of this work is to compare the quantitative performance of lung dosimetry in post-therapy Y90 gamma camera, SPECT, and PET imaging.

Materials and Methods: A modular phantom with a water-filled liver and 3 cm diameter tumor and two Styrofoam bead-filled lung inserts was imaged multiple times over the course of 2 weeks on a Siemens Intevo Bold SPECT/CT (7-minute planar and 25-minute SPECT), GE Discovery MI PET/CT (25-minute), and Siemens Quadra PET/CT (25-minute). The phantom contained initial activities resulting in 64 Gy to normal liver, 634 Gy to tumor, 12 Gy to right lung (6.7% LSF), and 26 Gy to left lung (13.2% LSF). Acquisition parameters followed clinical protocols (planar, SPECT, and DMI) or default vendor parameters (Quadra). For each image, an image scaling factor was derived from the known activity in the liver and the measured liver contour total signal. Individual lung activity was calculated from the product of lung contour total signals and the scaling factor. LSFs were calculated using the ratio of contour counts, and lung doses were calculated using the known lung masses of 440 g. For brevity, only results of a single acquisition per modality are reported herein.

Results: Calculated image scaling factors (Bq/count in planar and SPECT or Bq/Bq in PET) and errors in estimated LSFs and lung doses are tabulated. Overall, the DMI PET images had the lowest errors within ±6%, followed by underestimations of up to -12% by planar, -33% with SPECT, and -76% with Quadra PET. Given the measured liver activities agreed within 6% across both PET systems, the large errors in Quadra lung doses suggest optimization is still needed for lung activity quantification.

Conclusion: Initial results with high lung uptake phantoms suggest that doses can be estimated within 10% using SPECT or PET imaging but cross-modality harmonization is still needed. Protocol optimization with phantom data is underway to establish the minimum detectable lung microsphere deposition with each modality prior to performing inter-modality comparisons in patient data..