(Wednesday, 7/17/2019) 9:30 AM - 10:00 AM
Room: Exhibit Hall | Forum 5
Purpose: Expand the use of biomechanical model-based deformable image registration (DIR) available in commercial radiotherapy treatment planning system (TPS) by resolving organ overlap issue that limits the use of these tools. This study investigates solving this issue for the complex case of vertebral bodies to support the application of biomechanical modeling to resolve positional deformation associated with neck flexion for head and neck (HN) patients.
Methods: Two CT scans of 71 patients were retrospectively evaluated. Bony and soft tissue contours were automatically segmented. Neck flexion was estimated by a commercial biomechanical model-based DIR using C1 to C7 vertebra, mandible, skull and body contours to define boundary conditions. The biomechanical model requires no organ overlap and high quality meshes, therefore, an algorithm was developed to extract the organ surfaces of the TPS, correct the overlapping area using the Demons algorithm while preserving mesh topology, define the per organ rigid alignment using the iterative closest points method, and send the boundary conditions to the TPS. Accuracy of the biomechanical model to position the soft tissue, not considered in the model, was quantified.
Results: The completely automated workflow was successfully implemented within a scripting process of the TPS and allowed an estimation of the independent deformation induced by the neck flexion and mandible displacement only for all patients. The biomechanical model aligned the spatial position of the tumor, parotid and submandibular glands, spinal cord, and larynx but maintained the ability to quantify the volumetric response of the structures.
Conclusion: The algorithm automatically eliminates the organ overlap which prohibits the use of multi-organ biomechanical registration, enabling widespread use in the HN, abdomen, and pelvis. In this application, extracting independent deformations without deforming soft tissue, contrary to intensity-based DIR, may find interests for multimodal diagnostic images, evaluating treatment response, and mapping dose onto recurrence imaging.
Funding Support, Disclosures, and Conflict of Interest: Funding from RaySearch Laboratories
Keywords
Taxonomy
Not Applicable / None Entered.
Contact Email
