Gamma-index and dose-volume histograms (based on voxel dosimetry) to evaluate the predictive power of 99mTc-MAA SPECT maps in comparison with post-radioembolization 90Y PET maps

Radioembolization (RE) with
⁹⁰
Y microspheres has been recognized as an emerging treatment strategy for patients with liver tumors either primary or secondary. New tools for complex patient dosimetry, based on real absorbed dose distribution to the normal liver and tumor are under development by our research group at the Champalimaud Centre for the Unknown (CCU), Champalimaud Foundation (CF). Furthermore, we have been using the gamma-index quantitative approach to evaluate the degree of agreement between reference and target voxel-based dose distributions. In this research work, we quantitatively investigated the correlation between dose distributions calculated based on pre-treatment/reference Technetium-99m Macroaggregated Albumin (
^99m
Tc-MAA) Single Photon Emission Computed Tomography (SPECT) and post-treatment/target Yttrium-90 (
⁹⁰
Y) Positron Emission Tomography (PET) images, to evaluate the predictive power of
^99m
Tc-MAA SPECT maps in comparison with post-radioembolization
⁹⁰
Y PET maps. Voxel-based dose distributions were retrospectively calculated based on a total of 13 treatments from 10 patients submitted to RE with
⁹⁰
Y glass microspheres at CCU-CF. The dose distributions (reference and target) were co-registered to the same matrix resolution and a region extraction algorithm was applied to the entire data for each treatment so that only the liver was considered as the region of interest (ROI). In addition, each slice was normalized to the pixel with the maximum intensity value and a dose threshold was applied to the background level. The gamma-index test was then applied to all reference and target pairs of slices from each treatment image series (
^99m
Tc-MAA SPECT and
⁹⁰
Y PET), using the dose-difference (DD) and distance-to-agreement (DTA) tolerances. The computation based on 1766 dose maps (slices) included in this work, showed that the mean percentage of voxels passing gamma-index, for each patient, varies from 17% to 48%. Considering all the patients, the median of this variability is 35% and the mean value is 33% with a standard deviation of 10% In conclusion, the application of gamma-index tests to our data, with the tolerances considered (DD = 10% and DTA = 5mm), reveals good correlation between pre-treatment
^99m
Tc-MAA SPECT dose maps and post-treatment
⁹⁰
Y PET dose maps in one third of the voxels analyzed. Considering that the deposition of the microspheres (glass and MAA) and the dose maps are heterogeneous, these preliminary results are encouraging and led us to further consider the application of this gamma-index quantitative approach together with dose-volume histograms to investigate the additional predictive power of
^99m
Tc-MAA SPECT maps in comparison with post-therapy
⁹⁰
Y PET maps.