Room: Exhibit Hall | Forum 5
Purpose: To compensate the respiration motion of image-guided radiotherapy, a motion estimation method combining statistical motion modeling with thoraco-abdominal surface matching was proposed.
Methods: The motion modeling framework was built on the four-dimensional computed tomography (4D-CT) and the thoraco-abdominal surface acquired during radiotherapy procedure. The displacement vector field (DVF) from 4D-CT was utilized to construct a statistical motion model using the principal component analysis (PCA) technique. The patientâ€™s thoraco-abdominal surface was acquired by a depth camera in real time, then a similarity metric based on point cloud distance between pre- and intra-operative surfaces was constructed. Through maximizing the surface similarity metric, an internal DVF corresponding to current respiratory state can be obtained. Bayesian inference was applied to construct the metric to replace the surface similarity measurement. The 4D-CT dataset of ten patients was utilized for evaluation, the thoraco-abdominal surface from CT images was extracted to simulate intraoperative surface. The estimated DVF was calculated with the input simulated surface which was excluded from the training dataset for each patient.
Results: The motion model accuracy was evaluated from the differences between the estimated DVF and the corresponding ground truth DVF. The respiratory cycle in the 4D CT was divided into ten phases, ninety tests were performed on each of the three motion models. The DVF estimation errors for the three models were 0.60 Â± 0.27 mm (original model), and 0.58 Â± 0.30 mm (model with Bayesian inference), respectively.
Conclusion: The accuracy of the proposed motion modeling framework was acceptable, thus there is a good potential that applying our method into application for respiratory motion compensation.