TY - GEN
T1 - Modified simultaneous motion estimation and image reconstruction (m-SMEIR) for 4D-CBCT
AU - Zhao, Cong
AU - Zhong, Yuncheng
AU - Wang, Jing
AU - Jin, Mingwu
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/5/23
Y1 - 2018/5/23
N2 - Motion estimation using four-dimensional (4D) X-ray cone-beam computed tomography (CBCT) is important to achieve a precise radiation therapy of lung cancer patients. If the conventional three-dimensional (3D) reconstruction methods were used, the radiation dose would be much higher to obtain satisfying 4D images. Previously, we developed a simultaneous motion estimation and image reconstruction (SMEIR) method to reconstruct high-quality images and to obtain the motion model using a regular 3D CBCT scan. Due to non-convexity of the problem, the solution of SMEIR depends on the initial deformation vector fields (DVFs) and could be trapped in local optima. In this study, we develop a practical approach to alleviate this problem: First, each phase image is reconstructed using SMEIR with initial DVFs estimated from 3D reconstruction with total variation (TV) minimization. Then, DVFs are updated by applying the demons registration method on all SMEIR reconstructed phase images and used as the initial DVFs for the second round SMEIR. This modified SMEIR (m-SMEIR) method was tested using a CBCT simulation study of 4D NCAT digital phantom. The results demonstrate that m-SMEIR can yield better image quality than the SMEIR method. More importantly, m-SMEIR can produce much improved motion fields compared to SMEIR (47% improvement measured by the mean deviation from the true tumor motion), likely due to m-SMEIR's capability of jumping out of the local optimum.
AB - Motion estimation using four-dimensional (4D) X-ray cone-beam computed tomography (CBCT) is important to achieve a precise radiation therapy of lung cancer patients. If the conventional three-dimensional (3D) reconstruction methods were used, the radiation dose would be much higher to obtain satisfying 4D images. Previously, we developed a simultaneous motion estimation and image reconstruction (SMEIR) method to reconstruct high-quality images and to obtain the motion model using a regular 3D CBCT scan. Due to non-convexity of the problem, the solution of SMEIR depends on the initial deformation vector fields (DVFs) and could be trapped in local optima. In this study, we develop a practical approach to alleviate this problem: First, each phase image is reconstructed using SMEIR with initial DVFs estimated from 3D reconstruction with total variation (TV) minimization. Then, DVFs are updated by applying the demons registration method on all SMEIR reconstructed phase images and used as the initial DVFs for the second round SMEIR. This modified SMEIR (m-SMEIR) method was tested using a CBCT simulation study of 4D NCAT digital phantom. The results demonstrate that m-SMEIR can yield better image quality than the SMEIR method. More importantly, m-SMEIR can produce much improved motion fields compared to SMEIR (47% improvement measured by the mean deviation from the true tumor motion), likely due to m-SMEIR's capability of jumping out of the local optimum.
KW - 4D CBCT
KW - Local optimum
KW - Modified SMEIR (m-SMEIR)
KW - Motion fields
KW - SMEIR
UR - http://www.scopus.com/inward/record.url?scp=85048096256&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85048096256&partnerID=8YFLogxK
U2 - 10.1109/ISBI.2018.8363588
DO - 10.1109/ISBI.2018.8363588
M3 - Conference contribution
AN - SCOPUS:85048096256
T3 - Proceedings - International Symposium on Biomedical Imaging
SP - 340
EP - 343
BT - 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018
PB - IEEE Computer Society
T2 - 15th IEEE International Symposium on Biomedical Imaging, ISBI 2018
Y2 - 4 April 2018 through 7 April 2018
ER -