A 4-D maximum a posteriori (MAP) reconstruction method for improved dose volume histogram estimation

Objectives Moving from whole-organ to 3D dose estimation methods has the potential to improve the ability to predict biological response in targeted radionuclide therapy. 3D dose estimation methods are typically based on a time series of activity distributions obtained from independently-reconstructed 3D emission computed tomography (ECT) images. Noise and partial volume effects in these images degrade the estimated 3D dose-volume histograms. We have developed a new 4-D convergent MAP reconstruction method that provides noise reduction both spatially and across time points.

Methods To evaluate the method we simulated data using the 3D XCAT phantom to model patient anatomy and organ uptake based on In-111 Zevalin patient scans. Simulated analytic projection data, including Poisson noise, were generated representing realistic activity distributions at two time points. Images were reconstructed using both the proposed 4D method and OS-EM. Both reconstructions included compensation for image degrading effects. OS-EM images were optionally post-filtered to reduce noise. Images were evaluated in terms of the error in cumulative activity volume histograms, serving as a proxy for dose-volume histograms, for the liver and kidney. We also computed the sum of the absolute value of the difference between the true histogram and the ones obtained from the reconstructed images.

Results The 4D MAP algorithm was able to substantially reduce image noise and do so without the resolution loss inherent in post-filtering. As a result, it more faithfully reproduced the histograms and had a substantially lower L1 norm.

Conclusions The proposed 4D MAP reconstruction method has the potential to significantly reduce error in 3D dose estimates and thereby improve prediction of biological response. The proposed algorithm also has potential applications in dynamic imaging