Click here to


Are you sure ?

Yes, do it No, cancel

Independent Validation of Deformable Image Registration for MR-Guided Adaptive Radiotherapy for Head and Neck Cancers

B McDonald1*, A Ohrt2 , S Vedam3 , B Elgohari4 , J Yang5 , J Wang6 , C Fuller7 , K Brock8 , (1) ,Houston, TX, (2) ,Houston, TX, (3) UT MD Anderson Cancer Center, Houston, TX, (4) UT MD Anderson Cancer Center, Houston, TX, (5) MD Anderson Cancer Center, Houston, TX, (6) MD Anderson Cancer Center, Houston, TX, (7) MD Anderson Cancer Center, Houston, TX, (8) UT MD Anderson Cancer Center, Houston, TX


(Tuesday, 7/16/2019) 3:45 PM - 4:15 PM

Room: Exhibit Hall | Forum 6

Purpose: To evaluate the quality and reproducibility of the CT-to-MR and MR-to-MR deformable image registration (DIR) algorithms in Monaco for the head and neck region.

Methods: Five oropharyngeal cancer patients underwent weekly CT and MR imaging during radiotherapy. Contours of gross tumor volumes (GTVs) and 12 organs at risk were drawn on CT-sim and propagated onto MR-sim using rigid registration. Contours were propagated in Monaco using two DIR
methods: CT-sim to weekly MR (CT-to-MR) and MR-sim to weekly MR (MR-to-MR). Contours were evaluated against physician-approved contours and against each other using the Dice similarity coefficient (DSC) and mean distance to agreement (DTA). A Wilcoxon signed rank test was performed for each structure to compare the two algorithms. The reproducibility of each algorithm was assessed on week 4 images for two patients by manually misaligning the images and performing DIR three times with each algorithm. The three sets of contours for each algorithm were compared by DSC and DTA.

Results: When compared to physician-approved contours, no obvious trends in DSC or DTA over time were observed for either DIR method (i.e. registration quality did not degrade as time between the sim and registered image increased). The MR-to-MR algorithm performed significantly better than CT-to-MR for the brain, oral cavity, and primary GTV in both DSC and DTA (p<0.05). No significant difference was found for any other structure, except the submandibular gland showed significantly higher DSC for CT-to-MR. Comparing CT-to-MR directly to MR-to-MR, DSC and DTA values demonstrated considerable disagreement between the methods (DSC<0.8 for the GTVs, cochleae, mandible, spinal cord, and esophagus). Perfect reproducibility was observed for both algorithms (DSC of 1.0 and DTA of 0.0 for all structures).

Conclusion: Both DIR algorithms demonstrated acceptable quality and perfect reproducibility, although neither algorithm performed consistently better. Further evaluation is required before clinical implementation.


Deformation, Image-guided Therapy, Registration


IM/TH- Image registration : Multi-modality registration

Contact Email