(Monday, 7/30/2018) 9:30 AM - 10:00 AM
Room: Exhibit Hall | Forum 5
Purpose: Recent developments make it possible to perform CBCT IGRT for Gamma Knife SRS patients. A common workflow involves definition of targets and normal tissues on MR, followed by registration of the MR images to the pretreatment CBCT. Patient localization utilizing this methodology requires minimal registration uncertainty, as the procedure relies on submillimeter overall delivery accuracy. This work investigates the uncertainty associated with the MR to CBCT registration.
Methods: Anthropomorphic phantom studies are used to evaluate GammaPlan’s MR to CBCT registration accuracy, including geometric consistency of points and centroids of planar contours. Displacements found are compared to pre-measured geometric distortions for the MR scanner and phantom combination. Additionally, Dice similarity scoring is used to evaluate consistency for volumes. Points and contours are generated using threshold-based segmentation to mitigate intra/interobserver variability.
Results: Average registration point displacements are 0.3, 0.2, and 0.4mm for x, y, and z planes, respectively with total average 3D displacement of 0.5mm. Centroid displacements for 2D contours are 0.2, 0.3, and 0.4mm for x, y, and z planes, respectively with total average 3D displacement of 0.5mm. Total average displacement for all points and 2D contour centroids are within 0.1mm of pre-measured average geometric distortions for the positions of the points and contours. Dice similarity coefficients for the volumetric portion of the study ranged from 0.91-0.98 with an average value of 0.95.
Conclusion: Uncertainties associated with the rigid-frame-based system including but not limited to the inherent MR geometric distortion have collectively been shown to produce errors well in excess of 1mm. The total displacement due to MR to CBCT registration is well within the error reported for the rigid-frame-based system, and similar in magnitude to the MR distortion alone. The volumetric structure analysis gave Dice similarity scores well above the value recommended by the TG132 for registration accuracy.
Keywords
Not Applicable / None Entered.
Taxonomy
IM- Cone Beam CT: Registration
Contact Email
