Encrypted login | home

Program Information

Evaluation of On-Treatment Respiratory Motion Model Built From 3DCBCT Projections

no image available
B Liu

B Liu1*, Y Li1 , B Liang1 , R Wei1 , X Xu1 , B Guo1 , F Zhou1 , S Xu2 , H Xu3 , (1) Image Processing Center, Beihang University, Beijing, 100191 (2) Chinese PLA General Hospital, Beijing, 100853 (3) 302 Military Hospital, Beijing, 100039


SU-I-GPD-J-20 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: Motion model-based volumetric imaging during treatment has many useful applications. While the motion model built using 4DCT (MM-4DCT) may not reliably represent patient anatomy during treatment, constructing the model using 4DCBCT suffers from other shortcomings, such as excessive imaging dose. In this work, we evaluated a new method of building motion model using 3DCBCT projections (MM-3DCBCT).

Methods: One 4DCT and four 4DCBCT (4DCBCT1~4DCBCT4) images with a tumor located in left lung were simulated. 4DCBCT1 had the same anatomy with 4DCT. Compared with the 4DCT, the tumor baseline was shifted to the LR, AP and SI directions by 4 mm for 4DCBCT2~4DCBCT4 respectively. Thirty X-ray projections were simulated as the 3DCBCT projections for each phase of each 4DCBCT, and additional 30 real-time X-ray projections with random angles and breath phases were also simulated for each 4DCBCT. Two PCA-based motion models were constructed for each 4DCBCT. While one was MM-4DCT constructed using 4DCT, the other was MM-3DCBCT constructed via deformable registration of the end-exhale phase of the 4DCT to the 3DCBCT projections. Volumetric images were calculated from each real-time X-ray image based on these two models and the location of tumor centroid was detected. The location accuracy of the tumor centroid was compared for the two motion models.

Results: When using MM-4DCT, the location errors (X, Y, Z and 3D in mm) were (0.12±0.08, 0.86±0.52, 0.66±0.22, 1.12±0.49) , (2.12±0.13, 0.87±0.69, 0.60±0.58, 2.49±0.45), (0.13±0.09, 2.15±0.89, 0.54±0.26, 2.23±0.89) and (0.19±0.08, 0.63±0.63, 1.94±0.76, 2.11±0.82) for 4DCBCT1~4DCBCT4 respectively. When using MM-3DCBCT, the location errors were (0.14±0.09, 0.88±0.62, 0.67±0.25, 1.17±0.52), (0.93±0.21, 0.91±0.63, 0.80±0.42, 1.60±0.60), (0.09±0.06, 1.06±0.71, 0.56±0.32, 1.31±0.58) and (0.09±0.05, 0.75±0.58, 0.86±0.59, 1.18±0.77) for 4DCBCT1~4DCBCT4 respectively.

Conclusion: It is feasible to construct motion model using 3DCBCT projections. More accurate tumor localization can be achieved using MM-3DCBCT than MM-4DCT when there is shift of tumor base.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by the National Natural Science Foundation of China (61601012).

Contact Email: