SU-E-J-94: Geometric and Dosimetric Evaluation of Deformation Image Registration Algorithms Using Virtual Phantoms Generated From Patients with Lung Cancer
- The Cleveland Clinic, Cleveland, OH (United States)
- Philips, Fitchburg, WI (United States)
Purpose: To generate virtual phantoms with clinically relevant deformation and use them to objectively evaluate geometric and dosimetric uncertainties of deformable image registration (DIR) algorithms. Methods: Ten lung cancer patients undergoing adaptive 3DCRT planning were selected. For each patient, a pair of planning CT (pCT) and replanning CT (rCT) were used as the basis for virtual phantom generation. Manually adjusted meshes were created for selected ROIs (e.g. PTV, lungs, spinal cord, esophagus, and heart) on pCT and rCT. The mesh vertices were input into a thin-plate spline algorithm to generate a reference displacement vector field (DVF). The reference DVF was used to deform pCT to generate a simulated replanning CT (srCT) that was closely matched to rCT. Three DIR algorithms (Demons, B-Spline, and intensity-based) were applied to these ten virtual phantoms. The images, ROIs, and doses were mapped from pCT to srCT using the DVFs computed by these three DIRs and compared to those mapped using the reference DVF. Results: The average Dice coefficients for selected ROIs were from 0.85 to 0.96 for Demons, from 0.86 to 0.97 for intensity-based, and from 0.76 to 0.95 for B-Spline. The average Hausdorff distances for selected ROIs were from 2.2 to 5.4 mm for Demons, from 2.3 to 6.8 mm for intensity-based, and from 2.4 to 11.4 mm for B-Spline. The average absolute dose errors for selected ROIs were from 0.2 to 0.6 Gy for Demons, from 0.1 to 0.5 Gy for intensity-based, and from 0.5 to 1.5 Gy for B-Spline. Conclusion: Virtual phantoms were modeled after patients with lung cancer and were clinically relevant for adaptive radiotherapy treatment replanning. Virtual phantoms with known DVFs serve as references and can provide a fair comparison when evaluating different DIRs. Demons and intensity-based DIRs were shown to have smaller geometric and dosimetric uncertainties than B-Spline. Z Shen: None; K Bzdusek: an employee of Philips Healthcare; J Greskovich: None; P Xia: received research grants from Philips Healthcare and Siemens Healthcare.
- OSTI ID:
- 22494112
- Journal Information:
- Medical Physics, Journal Name: Medical Physics Journal Issue: 6 Vol. 42; ISSN 0094-2405; ISSN MPHYA6
- Country of Publication:
- United States
- Language:
- English
Similar Records
SU-E-J-119: Head-And-Neck Digital Phantoms for Geometric and Dosimetric Uncertainty Evaluation of CT-CBCT Deformable Image Registration
Analysis of deformable image registration accuracy using computational modeling
MO-C-17A-13: Uncertainty Evaluation of CT Image Deformable Registration for H and N Cancer Adaptive Radiotherapy
Journal Article
·
Mon Jun 15 00:00:00 EDT 2015
· Medical Physics
·
OSTI ID:22494132
Analysis of deformable image registration accuracy using computational modeling
Journal Article
·
Mon Mar 15 00:00:00 EDT 2010
· Medical Physics
·
OSTI ID:22096659
MO-C-17A-13: Uncertainty Evaluation of CT Image Deformable Registration for H and N Cancer Adaptive Radiotherapy
Journal Article
·
Sun Jun 15 00:00:00 EDT 2014
· Medical Physics
·
OSTI ID:22407766