Investigation of the accuracy of organ dose rate histogram estimates obtained from quantitative SPECT images

Objectives Taking into account dose-rate and spatial distribution of radiation dose has the potential to improve the accuracy of predicting biological response. However, dose rate estimates are usually based on activity distribution estimates obtained from emission computed tomography (ECT) imaging and are degraded by noise and limited spatial resolution. The goal was to investigate the effect of these degrading factors on the accuracy of dose-rate histograms.

Methods We used the realistic 3D NCAT phantom to model patient anatomy and 0-1 hr post injection organ activities based on I-131 Bexxar patient scans. Simulated projection data were generated using previously-validated Monte Carlo (MC) simulation techniques. The projection data had clinically realistic noise levels and modeled acquisition with a HEGP collimator. Images were reconstructed using OS-EM with model-based compensation for attenuation, scatter and the collimator-detector response. Dose-rate volume histograms were generated from SPECT images obtained with 10 and 25 iterations of 24 subsets using the 3D-RD MC dose estimation package. These histograms were compared with ones obtained from the true activity distribution.

Results There was significant error in the dose-rate histograms obtained from the SPECT images. The errors observed included both broadening of the histogram and a shift in the mean dose rate. We found that, especially for the liver, the histogram obtained with the smaller number of iterations was closer to the true histogram. These results indicate the need for partial volume compensation and noise regularization.

Conclusions Partial volume and noise in ECT images produce errors in dose-rate histograms. The magnitude and nature of the errors depend on factors such as organ size, imaging and reconstruction methods. Development of optimal techniques and methods is important in order to improve the ability of advanced dosimetric techniques to predict biological effects