Poster — Thur Eve — 14: Improving Tissue Segmentation for Monte Carlo Dose Calculation using DECT

Authors


Abstract

Purpose:

To improve Monte Carlo dose calculation accuracy through a new tissue segmentation technique with dual energy CT (DECT).

Methods:

Electron density (ED) and effective atomic number (EAN) can be extracted directly from DECT data with a stoichiometric calibration method. Images are acquired with Monte Carlo CT projections using the user code egs_cbct and reconstructed using an FDK backprojection algorithm. Calibration is performed using projections of a numerical RMI phantom. A weighted parameter algorithm then uses both EAN and ED to assign materials to voxels from DECT simulated images. This new method is compared to a standard tissue characterization from single energy CT (SECT) data using a segmented calibrated Hounsfield unit (HU) to ED curve. Both methods are compared to the reference numerical head phantom. Monte Carlo simulations on uniform phantoms of different tissues using dosxyz_nrc show discrepancies in depth-dose distributions.

Results:

Both SECT and DECT segmentation methods show similar performance assigning soft tissues. Performance is however improved with DECT in regions with higher density, such as bones, where it assigns materials correctly 8% more often than segmentation with SECT, considering the same set of tissues and simulated clinical CT images, i.e. including noise and reconstruction artifacts. Furthermore, Monte Carlo results indicate that kV photon beam depth-dose distributions can double between two tissues of density higher than muscle.

Conclusions:

A direct acquisition of ED and the added information of EAN with DECT data improves tissue segmentation and increases the accuracy of Monte Carlo dose calculation in kV photon beams.

Ancillary