Do Bulk Density Assignment method or Atlas method provide a better basis for an MRI-only
workflow?

Dose distributions based on sCTbda and on sCTa were compared to clinical dose distributions
based on the original planning CT in terms of clinically relevant DVH points and in terms
of 2D gamma analysis.
Three different approaches to generate sCTbda were evaluated in this thesis. For the sCTbda CTinfo
method, tissue classes in the original CT were segmented and assigned to approximated densities.
The purpose of this method was to evaluate the influence of a rough density estimation on
the accuracy of the dose distribution. In the sCTbda MRinfo method, the tissue class segmentation
was based on the MRI scans, both T1- and T2- weighted. Thus, this method was used to
evaluate MRI delineation uncertainty in addition to a rough electron density estimation. Dose
distribution based on sCTbda CTinfo showed slightly better agreement with the original dose
distribution than both sCTbda MRinfo. This was expected as sCTbda CTinfo introduces less
variation to the real CT geometry. As sCTbda CTinfo relies on CT information, its implementation
to establish an MRI-only workflow would not be viable. This method was only used to
assess the performance of tissue class delineation when using MRI. For both T1- and T2-MRinfo
sCTbda as well as for sCTbda CTinfo no clinically significant differences were found.
In addition to Bulk Density Assignment methods, Atlas based methods were investigated for the
generation of sCTs to determine which method is the best guess for geometry and electron density
estimation. Similar dosimetric results were obtained for sCTa generated with and without
bone refinement. Dose distributions based on sCTa showed slightly better PTV coverage and
a more homogeneous gamma map with lower gamma values than sCTbda MRinfo. As for the
Atlas methods a one-to-one estimation for each electron density voxel value is assigned, a greater
similarity with the original dose distributions is expected to be obtained. However, differences
seen between sCTa and sCTbda were not clinically significant.

http://repositorio.ul.pt/bitstream/10451/20395/1/ulfc114379_...

--------

In this thesis, we propose to evaluate two of the most promising MRI calibration approaches
for the generation of synthetic CT’s (sCT): The Manual Bulk Density Assignment (sCTbda) and
the Multi Atlas Registration (sCTa) schemes. These approaches were evaluated on patient data
sets for which CT and MR images, acquired at the same day and under the same conditions,
exist for head and neck (H&N) and prostate sites. Clinical Volumetric Modulated Arc Therapy
(VMAT) plans for the H&N patients and Intensity Modulated Radiation Therapy (IMRT) plans
for the prostate patients were re-calculated on both sCTbda and sCTa generated. - http://repositorio.ul.pt/bitstream/10451/20395/1/ulfc114379_...