Encrypted login | home

Program Information

A Non-Rigid Point Matching Method for Accurate Bladder Dose Summation in Cervical Cancer HDR Brachytherapy

no image available
X Zhen

H Chen1 , X Zhen1*, Z Zhong2 , A Pompos3 , H Yan3 , L Zhou1 , S Jiang3 , X Gu3 , (1) Southern Medical University, Guangzhou, Guangdong,510515 (2) The University of Texas at Dallas, Department of Computer Science, TX, (3) UT Southwestern Medical Center, Dallas, TX


SU-F-BRF-9 Sunday 4:00PM - 6:00PM Room: Ballroom F

Purpose: To propose and validate a deformable point matching scheme for surface deformation to facilitate accurate bladder dose summation for fractionated HDR cervical cancer treatment.
Method: A deformable point matching scheme based on the thin plate spline robust point matching (TPS-RPM) algorithm is proposed for bladder surface registration. The surface of bladders segmented from fractional CT images is extracted and discretized with triangular surface mesh. Deformation between the two bladder surfaces are obtained by matching the two meshes' vertices via the TPS-RPM algorithm, and the deformation vector fields (DVFs) characteristic of this deformation is estimated by B-spline approximation. Numerically, the algorithm is quantitatively compared with the Demons algorithm using five clinical cervical cancer cases by several metrics: vertex-to-vertex distance (VVD), Hausdorff distance (HD), percent error (PE), and conformity index (CI). Experimentally, the algorithm is validated on a balloon phantom with 12 surface fiducial markers. The balloon is inflated with different amount of water, and the displacement of fiducial markers is benchmarked as ground truth to study TPS-RPM calculated DVFs' accuracy.
Results: In numerical evaluation, the mean VVD is 3.7(±2.0) mm after Demons, and 1.3(±0.9) mm after TPS-RPM. The mean HD is 14.4 mm after Demons, and 5.3mm after TPS-RPM. The mean PE is 101.7% after Demons and decreases to 18.7% after TPS-RPM. The mean CI is 0.63 after Demons, and increases to 0.90 after TPS-RPM. In the phantom study, the mean Euclidean distance of the fiducials is 7.4±3.0mm and 4.2±1.8mm after Demons and TPS-RPM, respectively.
Conclusions: The bladder wall deformation is more accurate using the feature-based TPS-RPM algorithm than the intensity-based Demons algorithm, indicating that TPS-RPM has the potential for accurate bladder dose deformation and dose summation for multi-fractional cervical HDR brachytherapy.

Funding Support, Disclosures, and Conflict of Interest: This work is supported in part by the National Natural ScienceFoundation of China (no 30970866 and no 81301940).

Contact Email: