Physics-based high-resolution head and neck biomechanical models
Systems and methods are shown for developing physics-based high resolution biomechanical head and neck deformable models for generating ground-truth deformations that can be used for validating both image registration and adaptive RT frameworks.
1. An apparatus for deformable image registration (DIR) of a target anatomy, the apparatus comprising:
(a) a computer processor; and
(b) a non-transitory computer-readable memory storing instructions executable by the computer processor;
(c) wherein said instructions, when executed by the computer processor, perform steps comprising:
(i) generating a high-resolution deformable biomechanical model from CT data acquired from a plurality of patient CT images of the target anatomy;
(ii) simulating one or more deformations within said biomechanical model, the one or more deformations representing a range of inter-fraction patient posture changes and physiological regression; and
(iii) generating modified CT images representing a deformed state of the target anatomy;
(iv) wherein the target anatomy comprises a head and neck region of the patient and the biomechanical model is generated using one or more inter-fraction kVCT datasets and the corresponding structure contours of the head and neck region;
(v) wherein generating a high-resolution deformable biomechanical model comprises:
assigning voxels within the CT images to corresponding single structures as a function of contoured vertices within the CT images;
generating series of connected mass elements with associated mass-spring-damping (MSD) connections, each mass element being located at a center of each voxel;
connecting one or more mass elements with other mass elements via a spring damper formulation to ensure deformation of the one or mass elements in a specified physiological manner;
(vi) wherein bony anatomy within the CT images is modeled as rigid body and muscle and soft tissue structures are modeled as MSD models with elastic material properties that corresponded to the underlying contoured anatomies;
(vii) wherein simulating the one or more deformations comprises computing the biomechanical head and neck model deformations and actuating the biomechanical head and neck model to represent posture changes and physiological changes;
(viii) wherein computing the biomechanical head and neck model deformations comprises:
calculating internal corrective forces on each mass element as a summation of tensile spring force, shear spring force, and a dashpot damping force; and
updating the locations of each mass element from a previous iteration.
2. The apparatus of claim 1 , wherein voxels inside a given 3D contour boundary are assigned and clustered using a GPU-based algorithm.
3. The apparatus of claim 1 , wherein head and neck deformations are configured to be actuated by user-defined rigid transformations of one or more skeletal structures.
4. The apparatus of claim 3 :
wherein soft-tissue deformation for a given skeletal actuation is performed using an implicit Euler integration; and
wherein each iteration of the Euler integration is split into individual iterative steps for the muscle structures and remaining soft tissue structures.
5. The apparatus of claim 4 :
wherein the one or more skeletal structures are modeled as rigid body; and
wherein the muscle and soft tissue structures are modeled as mass spring models with elastic material properties that correspond to underlying contoured anatomies.
6. The apparatus of claim 4 , wherein within a given muscle structure, the voxels are classified using a uniform grid, and a normalized mass is assigned to each voxel based on its Hounsfield number.
7. The apparatus of claim 5 , wherein posture changes are simulated by articulating the one or more skeletal structures and enabling the soft structures to deform accordingly.
8. The apparatus of claim 1 :
wherein physiological regression comprises a tumor regression; and
wherein said physiological regression is simulated by reducing a target volume and enabling surrounding soft tissue structures to deform accordingly.
9. The apparatus of claim 1 :
wherein the modified CT images comprise a 3D contour boundary having one or more of inconsistencies and gaps; and
wherein voxels inside a given 3D contour boundary are clustered to form a deformable volumetric structure to account for the one or more of inconsistencies and gaps.
10. The apparatus of claim 1 :
wherein the processor comprises a graphics processing unit (GPU);
wherein the simulations of posture changes and physiological regressions are performed at interactive speeds.
11. A method for deformable image registration (DIR) of a target anatomy, the method comprising:
(a) generating a high-resolution deformable biomechanical model from CT data acquired from a plurality of patient CT images of the target anatomy;
(b) simulating one or more deformations within said biomechanical model, the one or more deformations representing a range of inter-fraction patient posture changes and physiological regression; and
(c) generating modified CT images representing a deformed state of the target anatomy;
(d) wherein the target anatomy comprises a head and neck region of the patient and the biomechanical model is generated using one or more inter-fraction kVCT datasets and the corresponding structure contours of the head and neck region;
(e) wherein generating a high-resolution deformable biomechanical model comprises:
(i) assigning voxels within the CT images to corresponding single structures as a function of contoured vertices within the CT images;
generating series of connected mass elements with associated mass-spring-damping (MSD) connections, each mass element being located at a center of each voxel;
(ii) connecting one or more mass elements with other mass elements via a spring damper formulation to ensure deformation of the one or mass elements in a specified physiological manner;
(f) wherein bony anatomy within the CT images is modeled as rigid body and muscle and soft tissue structures are modeled as MSD models with elastic material properties that corresponded to the underlying contoured anatomies;
(g) wherein simulating the one or more deformations comprises computing the biomechanical head and neck model deformations and actuating the biomechanical head and neck model to represent posture changes and physiological changes;
(h) wherein computing the biomechanical head and neck model deformations comprises:
(i) calculating internal corrective forces on each mass element as a summation of tensile spring force, shear spring force, and a dashpot damping force; and
(ii) updating the locations of each mass element from a previous iteration;
(i) wherein said method is performed by instructions executable by a computer processor and are stored on a non-transitory computer-readable memory.
12. The method of claim 11 , wherein voxels inside a given 3D contour boundary are assigned and clustered using a GPU-based algorithm.
13. The method of claim 11 , wherein head and neck deformations are configured to be actuated by user-defined rigid transformations of one or more skeletal structures.
14. The method of claim 13 :
wherein soft-tissue deformation for a given skeletal actuation is performed using an implicit Euler integration; and
wherein each iteration of the Euler integration is split into individual iterative steps for the muscle structures and remaining soft tissue structures.
15. The method of claim 14 :
wherein the one or more skeletal structures are modeled as rigid body; and
wherein the muscle and soft tissue structures are modeled as mass spring models with elastic material properties that corresponded to underlying contoured anatomies.
16. The method of claim 14 , wherein within a given muscle structure, the voxels are classified using a uniform grid, and a normalized mass is assigned to each voxel based on its Hounsfield number.
17. The method of claim 13 , wherein posture changes are simulated by articulating the skeletal structure and enabling the soft structures to deform accordingly.
18. The method of claim 11 :
wherein physiological regression comprises a tumor regression; and
wherein said physiological regression is simulated by reducing the target volume and enabling surrounding soft structures to deform accordingly.
19. The method of claim 11 :
wherein the modified CT images comprise a 3D contour boundary having one or more of inconsistencies and gaps; and
wherein voxels inside a given 3D contour boundary are clustered to form a deformable volumetric structure to account for the one or more of inconsistencies and gaps.
20. The method of claim 11 :
wherein the processor comprises a graphics processing unit (GPU);
wherein the simulations of posture changes and physiological regressions are performed at interactive speeds.
21. An apparatus for deformable image registration (DIR) of a target anatomy, the apparatus comprising:
(a) a computer processor; and
(b) a non-transitory computer-readable memory storing instructions executable by the computer processor;
(c) wherein said instructions, when executed by the computer processor, perform steps comprising:
(i) generating a high-resolution deformable biomechanical model from CT data acquired from a plurality of patient CT images of the target anatomy;
(ii) simulating one or more deformations within said biomechanical model, the one or more deformations representing a range of inter-fraction patient posture changes and physiological regression; and
(iii) generating modified CT images representing a deformed state of the target anatomy;
(iv) wherein the target anatomy comprises a head and neck region of the patient and the biomechanical model is generated using one or more inter-fraction kVCT datasets and the corresponding structure contours of the head and neck region;
(v) wherein generating a high-resolution deformable biomechanical model comprises:
assigning voxels within the CT images to corresponding single structures as a function of contoured vertices within the CT images;
generating series of connected mass elements with associated mass-spring-damping (MSD) connections, each mass element being located at a center of each voxel;
connecting one or more mass elements with other mass elements via a spring damper formulation to ensure deformation of the one or mass elements in a specified physiological manner;
(vi) wherein bony anatomy within the CT images is modeled as rigid body and muscle and soft tissue structures are modeled as MSD models with elastic material properties that corresponded to the underlying contoured anatomies;
(vii) wherein simulating the one or more deformations comprises computing the biomechanical head and neck model deformations and actuating the biomechanical head and neck model to represent posture changes and physiological changes;
(viii) wherein head and neck deformations are configured to be actuated by user-defined rigid transformations of one or more skeletal structures;
(viii) wherein soft-tissue deformation for a given skeletal actuation is performed using an implicit Euler integration; and
(ix) wherein each iteration of the Euler integration is split into individual iterative steps for the muscle structures and remaining soft tissue structures.
22. An apparatus for deformable image registration (DIR) of a target anatomy, the apparatus comprising:
(a) a computer processor; and
(b) a non-transitory computer-readable memory storing instructions executable by the computer processor;
(c) wherein said instructions, when executed by the computer processor, perform steps comprising:
(i) generating a high-resolution deformable biomechanical model from CT data acquired from a plurality of patient CT images of the target anatomy;
(ii) simulating one or more deformations within said biomechanical model, the one or more deformations representing a range of inter-fraction patient posture changes and physiological regression; and
(iii) generating modified CT images representing a deformed state of the target anatomy;
(iv) wherein the target anatomy comprises a head and neck region of the patient and the biomechanical model is generated using one or more inter-fraction kVCT datasets and the corresponding structure contours of the head and neck region;
(v) wherein generating a high-resolution deformable biomechanical model comprises:
assigning voxels within the CT images to corresponding single structures as a function of contoured vertices within the CT images;
generating series of connected mass elements with associated mass-spring-damping (MSD) connections, each mass element being located at a center of each voxel;
connecting one or more mass elements with other mass elements via a spring damper formulation to ensure deformation of the one or mass elements in a specified physiological manner;
(vi) wherein bony anatomy within the CT images is modeled as rigid body and muscle and soft tissue structures are modeled as MSD models with elastic material properties that corresponded to the underlying contoured anatomies;
(vii) wherein simulating the one or more deformations comprises computing the biomechanical head and neck model deformations and actuating the biomechanical head and neck model to represent posture changes and physiological changes;
(viii) wherein physiological regression comprises a tumor regression; and
(ix) wherein said physiological regression is simulated by reducing a target volume and enabling surrounding soft tissue structures to deform accordingly.
23. A method for deformable image registration (DIR) of a target anatomy, the method comprising:
(a) generating a high-resolution deformable biomechanical model from CT data acquired from a plurality of patient CT images of the target anatomy;
(b) simulating one or more deformations within said biomechanical model, the one or more deformations representing a range of inter-fraction patient posture changes and physiological regression; and
(c) generating modified CT images representing a deformed state of the target anatomy;
(d) wherein the target anatomy comprises a head and neck region of the patient and the biomechanical model is generated using one or more inter-fraction kVCT datasets and the corresponding structure contours of the head and neck region;
(e) wherein generating a high-resolution deformable biomechanical model comprises:
(i) assigning voxels within the CT images to corresponding single structures as a function of contoured vertices within the CT images;
(ii) generating series of connected mass elements with associated mass-spring-damping (MSD) connections, each mass element being located at a center of each voxel;
(iii) connecting one or more mass elements with other mass elements via a spring damper formulation to ensure deformation of the one or mass elements in a specified physiological manner;
(f) wherein bony anatomy within the CT images is modeled as rigid body and muscle and soft tissue structures are modeled as MSD models with elastic material properties that corresponded to the underlying contoured anatomies;
(g) wherein simulating the one or more deformations comprises computing the biomechanical head and neck model deformations and actuating the biomechanical head and neck model to represent posture changes and physiological changes;
(h) wherein head and neck deformations are configured to be actuated by user-defined rigid transformations of one or more skeletal structures;
(i) wherein soft-tissue deformation for a given skeletal actuation is performed using an implicit Euler integration; and
(j) wherein each iteration of the Euler integration is split into individual iterative steps for the muscle structures and remaining soft tissue structures;
(k) wherein said method is performed by instructions executable by a computer processor and are stored on a non-transitory computer-readable memory.
24. A method for deformable image registration (DIR) of a target anatomy, the method comprising:
(a) generating a high-resolution deformable biomechanical model from CT data acquired from a plurality of patient CT images of the target anatomy;
(b) simulating one or more deformations within said biomechanical model, the one or more deformations representing a range of inter-fraction patient posture changes and physiological regression; and
(c) generating modified CT images representing a deformed state of the target anatomy;
(d) wherein the target anatomy comprises a head and neck region of the patient and the biomechanical model is generated using one or more inter-fraction kVCT datasets and the corresponding structure contours of the head and neck region;
(e) wherein generating a high-resolution deformable biomechanical model comprises:
(i) assigning voxels within the CT images to corresponding single structures as a function of contoured vertices within the CT images;
(ii) generating series of connected mass elements with associated mass-spring-damping (MSD) connections, each mass element being located at a center of each voxel;
(iii) connecting one or more mass elements with other mass elements via a spring damper formulation to ensure deformation of the one or mass elements in a specified physiological manner;
(f) wherein bony anatomy within the CT images is modeled as rigid body and muscle and soft tissue structures are modeled as MSD models with elastic material properties that corresponded to the underlying contoured anatomies;
(h) wherein simulating the one or more deformations comprises computing the biomechanical head and neck model deformations and actuating the biomechanical head and neck model to represent posture changes and physiological changes;
(i) wherein physiological regression comprises a tumor regression; and
(j) wherein said physiological regression is simulated by reducing a target volume and enabling surrounding soft tissue structures to deform accordingly;
(k) wherein said method is performed by instructions executable by a computer processor and are stored on a non-transitory computer-readable memory.