Brain MR-Only Using a Hybrid CT

Abstract

The accuracy of synthetic CT images, generated for the MR-only workflow, can be challenged by tissue segmentation errors as well as improper electron density assignment for a given patient. At some point during their diagnostic workup most brain tumor patients are imaged with CT. Furthermore, many centers perform brain MR simulation using diagnostic RF coils and then apply rigid registration to align the MR images to planning CT images. We demonstrate here that the inverse approach (applying rigid registration to align diagnostic CT images to planning MR images) can be used to generate hybrid CTs suitable for the MR-only workflow in brain. Twelve brain tumor patients, immobilized in treatment masks, were imaged on a Siemens 3T using a clinical MR simulation protocol. A 3D T1 image was selected as the reference dataset. Hybrid CTs were constructed in MIM (MIM Software) as follows: 1) DICOM screeners verified 3D T1 integrity (axial slices, WFS<1mm, 3D GNL correction), 2) 3D T1 images were standardized, 3) non-contrast diagnostic CTs were upsampled and rigidly registered to 3D T1 images, 4) an external patient contour generated from the 3D T1 was filled with HU value of water, 5) voxels within the external contour were replaced by diagnostic CT voxels wherever possible. The hybrid CTs were then transferred to Monaco (Elekta) where two dosimetric comparisons were performed: 1) original VMAT plans were templated and recalculated on the hybrid CTs, and 2) VMAT plans were reoptimized on the hybrid CTs and then recalculated on the planning CTs. The Gammex 467 phantom was imaged with the brain protocol on diagnostic CT scanners to verify CTED conversion agreement with the clinical CTED file in Monaco. Dose distributions were compared using 3D gamma analysis in 3DVH (Sun Nuclear). D95PTV and OAR dose-volume constraints were also compared. All patients had non-contrast diagnostic CTs prior to MR simulation. No significant difference in CTED assignment existed from the diagnostic CT scanners. As shown in the Table, 3D dose distribution matching rates were within 1.6% and D95PTV percent differences were within 1.2%. OAR dose-volume parameters were consistent between plans calculated on planning and hybrid CTs. These results demonstrate that, by using an actual CT of the patient, hybrid CTs eliminate a majority of segmentation and density assignment uncertainties that can affect the accuracy of dose calculation. DRRs were successfully generated from the hybrid CTs and transferred to Mosaiq (Elekta) and XVI (Elekta). Standardized 3D T1 images were successfully transferred to XVI and loaded as reference images for IGRT. Hybrid CTs, suitable for the brain MR-only workflow, can be constructed from rigid registration of diagnostic CTs to planning MR images. No additional scan time or custom sequences are required.Abstract 3688Original Plan Recalculated on Hybrid CTReoptimized Plan Recalculated on Planning CT3D Match Rate; D95PTV %Diff3D Match Rate; D95PTV %Diff99.1%; 0%98.4%; -1.2% Open table in a new tab