Physics-Based Patient-Specific Respiration Modeling For 4d Treatment Planning
C. Shi,J. Eom,D. Vines,S. De,G.X. Xu
DOI: https://doi.org/10.1016/j.ijrobp.2009.07.1481
2009-01-01
Abstract:Purpose/Objective(s)To develop a tool for IGRT/IMRT of lung cancer by using physical-based patient-specific dynamic lung model from 4DCT data and pressure-volume relationship.Materials/MethodsA finite element (FE) model of lung motion was developed from 4DCT images of one patient with a tumor in the right lung. The 4DCT images were sorted into 10 phases and region of interests (ROIs) of each phase were drawn by one physician. The ROIs were extracted and further processed into a FE meshes using the computer-aided design (CAD) surface reconstruction approach. The pressure volume (PV) curves were applied to deform the ROIs meshes into 4D. The PV curve represents the elastic properties of the lung including the nonlinearity which makes the lung stiffer at higher volumes and hysteresis between inflation and deflation. Nonlinearity from the soft tissue material of the lung and the contact condition from the pleural sliding were appropriately treated. Features points selected from 4D CT were used as the deformation benchmarks.ResultsThe FE model had 66704 tetrahedral elements with 3 independent components and 55947 unknown variables in total. An inspection of the overall lung motion pattern as driven by the PV curve shows that the modeling is physiologically realistic. Though the tumor motion in the upper lobe of the lung is relatively small in comparison with maximum displacement of lung (1.13 cm in SI direction), a good agreement between the computed and those from the 4DCT is demonstrated. The average tumor motions in the upper lobe are 1.47 (SD=0.98) mm in the LR direction, 2.33 (1.016) mm in the SI direction, and 1.95 (1.01) mm in the AP direction. Specific to the lung tumor tracking and prediction, existing models have their own pros and cons. Image (pixel/voxel) based approach seems to be computationally more efficient but are incapable of capturing accurate organ deformations. Physics based models, as shown in this study, are capable of tracking and predicting the lung deformations to the precisions necessary for the clinical needs, however, it is computationally expensive. In particular, this study suggests that the use PV curve data to model the respiration has the following benefits. The PV relationship in the simulation enables a researcher to apply the physiologically correct boundary condition and to increase the accuracy of prediction. Temporal lung and tumor motions afford the opportunity to simulate the specific moment during a full breathing cycle, which will open the door for advanced radiation therapy.ConclusionsA physics-based, patient-specific dynamic respiration model has been developed from a real patient 4D CT data and PV curves. The realistic physiology allows the model to exhibit accurate boundary conditions needed for many advanced treatment methods for the lung cancer. Purpose/Objective(s)To develop a tool for IGRT/IMRT of lung cancer by using physical-based patient-specific dynamic lung model from 4DCT data and pressure-volume relationship. To develop a tool for IGRT/IMRT of lung cancer by using physical-based patient-specific dynamic lung model from 4DCT data and pressure-volume relationship. Materials/MethodsA finite element (FE) model of lung motion was developed from 4DCT images of one patient with a tumor in the right lung. The 4DCT images were sorted into 10 phases and region of interests (ROIs) of each phase were drawn by one physician. The ROIs were extracted and further processed into a FE meshes using the computer-aided design (CAD) surface reconstruction approach. The pressure volume (PV) curves were applied to deform the ROIs meshes into 4D. The PV curve represents the elastic properties of the lung including the nonlinearity which makes the lung stiffer at higher volumes and hysteresis between inflation and deflation. Nonlinearity from the soft tissue material of the lung and the contact condition from the pleural sliding were appropriately treated. Features points selected from 4D CT were used as the deformation benchmarks. A finite element (FE) model of lung motion was developed from 4DCT images of one patient with a tumor in the right lung. The 4DCT images were sorted into 10 phases and region of interests (ROIs) of each phase were drawn by one physician. The ROIs were extracted and further processed into a FE meshes using the computer-aided design (CAD) surface reconstruction approach. The pressure volume (PV) curves were applied to deform the ROIs meshes into 4D. The PV curve represents the elastic properties of the lung including the nonlinearity which makes the lung stiffer at higher volumes and hysteresis between inflation and deflation. Nonlinearity from the soft tissue material of the lung and the contact condition from the pleural sliding were appropriately treated. Features points selected from 4D CT were used as the deformation benchmarks. ResultsThe FE model had 66704 tetrahedral elements with 3 independent components and 55947 unknown variables in total. An inspection of the overall lung motion pattern as driven by the PV curve shows that the modeling is physiologically realistic. Though the tumor motion in the upper lobe of the lung is relatively small in comparison with maximum displacement of lung (1.13 cm in SI direction), a good agreement between the computed and those from the 4DCT is demonstrated. The average tumor motions in the upper lobe are 1.47 (SD=0.98) mm in the LR direction, 2.33 (1.016) mm in the SI direction, and 1.95 (1.01) mm in the AP direction. Specific to the lung tumor tracking and prediction, existing models have their own pros and cons. Image (pixel/voxel) based approach seems to be computationally more efficient but are incapable of capturing accurate organ deformations. Physics based models, as shown in this study, are capable of tracking and predicting the lung deformations to the precisions necessary for the clinical needs, however, it is computationally expensive. In particular, this study suggests that the use PV curve data to model the respiration has the following benefits. The PV relationship in the simulation enables a researcher to apply the physiologically correct boundary condition and to increase the accuracy of prediction. Temporal lung and tumor motions afford the opportunity to simulate the specific moment during a full breathing cycle, which will open the door for advanced radiation therapy. The FE model had 66704 tetrahedral elements with 3 independent components and 55947 unknown variables in total. An inspection of the overall lung motion pattern as driven by the PV curve shows that the modeling is physiologically realistic. Though the tumor motion in the upper lobe of the lung is relatively small in comparison with maximum displacement of lung (1.13 cm in SI direction), a good agreement between the computed and those from the 4DCT is demonstrated. The average tumor motions in the upper lobe are 1.47 (SD=0.98) mm in the LR direction, 2.33 (1.016) mm in the SI direction, and 1.95 (1.01) mm in the AP direction. Specific to the lung tumor tracking and prediction, existing models have their own pros and cons. Image (pixel/voxel) based approach seems to be computationally more efficient but are incapable of capturing accurate organ deformations. Physics based models, as shown in this study, are capable of tracking and predicting the lung deformations to the precisions necessary for the clinical needs, however, it is computationally expensive. In particular, this study suggests that the use PV curve data to model the respiration has the following benefits. The PV relationship in the simulation enables a researcher to apply the physiologically correct boundary condition and to increase the accuracy of prediction. Temporal lung and tumor motions afford the opportunity to simulate the specific moment during a full breathing cycle, which will open the door for advanced radiation therapy. ConclusionsA physics-based, patient-specific dynamic respiration model has been developed from a real patient 4D CT data and PV curves. The realistic physiology allows the model to exhibit accurate boundary conditions needed for many advanced treatment methods for the lung cancer. A physics-based, patient-specific dynamic respiration model has been developed from a real patient 4D CT data and PV curves. The realistic physiology allows the model to exhibit accurate boundary conditions needed for many advanced treatment methods for the lung cancer.