Radiation Physics

PV QA 3 - Poster Viewing Q&A 3

TU_21_3322 - Evaluation of a Novel Hybrid Deformable Registration Algorithm for CBCT to CBCT Deformation in Prostate Studies

Tuesday, October 23
1:00 PM - 2:30 PM
Location: Innovation Hub, Exhibit Hall 3

Evaluation of a Novel Hybrid Deformable Registration Algorithm for CBCT to CBCT Deformation in Prostate Studies
S. Pirozzi1, N. Lamba2, A. Kruzer1, and A. S. Nelson1; 1MIM Software, Inc., Cleveland, OH, 2MIM Software, Cleveland, OH

Purpose/Objective(s): Previous studies have shown automatic deformation methods utilizing CBCT images have the potential to be incorporated into an automated prostate adaptive radiation therapy workflow in order to track daily delivered dose (and dose accumulated over time) to targets and organs-at-risk. Our goal in this current work is to improve the accuracy of CBCT to CBCT registration by introducing a hybrid deformation method and evaluate its accuracy as compared to a standard intensity based deformation method and manual contouring for prostate CBCT.

Materials/Methods: Sixty CBCTs were selected from nine patients with prostate cancer across multiple institutions. Contours were manually defined on each CBCT scan by a radiation oncologist for the bladder, rectum, and prostate. The Day 1 CBCT was deformed to each subsequent CBCT using two methods, including a normalized intensity based deformable registration algorithm (Intensity), and a novel hybrid contour-based algorithm (Hybrid). The Hybrid algorithm utilizes an image matching metric which minimizes both intensity differences between the two images and surface differences between corresponding contours defined on the two images. Dice coefficients were calculated for the bladder, rectum, and prostate comparing the contours generated by the two deformable methods to the contours generated by manual contouring. Paired t-tests were performed to determine the best automatic deformable image registration method. Table 1: Average Dice Similarity Coefficient
Structure Intensity Hybrid
Bladder 0.75 ± 0.12 (0.36 - 0.93) 0.91± 0.04 (0.77 - 0.96)
Prostate 0.74 ± 0.11 (0.40 - 0.93) 0.85 ± 0.06 (0.70 - 0.95)
Rectum 0.71 ± 0.09 (0.45 - 0.86) 0.85 ± 0.06 (0.54 - 0.93)

Results: Overall, utilizing the combined intensity plus contour-based registration improved average Dice scores across all patients for all structures from 0.73 +/- 0.11 (Intensity) to 0.87 ± 0.06 (Hybrid). Paired t-test results show that the Hybrid method was significantly better than the Intensity method (p < 0.001 for all contours).

Conclusion: While both Intensity and Hybrid deformable image registration methods showed good accuracy compared to manual contours in terms of Dice Coefficients (0.73 to 0.87), Hybrid was the most accurate method. We hypothesize that using this new method of deformation will also lead to more accurate dose transfer and accumulation due to the combination of both intensity information and contour boundaries to guide the deformation. This deformable image registration method has the potential to be incorporated into an automated prostate adaptive radiation therapy workflow for dose deformation in tracking accumulated dose.

Author Disclosure: S. Pirozzi: None. N. Lamba: None. A. Kruzer: None. A.S. Nelson: None.

Send Email for Sara Pirozzi


Assets

TU_21_3322 - Evaluation of a Novel Hybrid Deformable Registration Algorithm for CBCT to CBCT Deformation in Prostate Studies



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Evaluation of a Novel Hybrid Deformable Registration Algorithm for CBCT to CBCT Deformation in Prostate Studies