| dc.description.abstract | Surgical and interventional procedures such as resection, ablation, and biopsy depend on knowledge of the intraoperative positions of lesions and other relevant structures. However, preoperative chemotherapeutic regimens can impair intraoperative detection of tumors due to radiographic and sonographic disappearance of malignancy, arising from a gap between complete imaging and pathologic response. Furthermore, a shift towards minimally invasive and laparoscopic techniques diminishes the ability to detect, visualize, and manipulate intraoperative anatomy. Image guidance aims to overcome these barriers by providing an accurate reference for the patient anatomy during intervention, to improve the ability to localize and target designated structures and navigate around areas of complicative concern. Yet, in the context of therapy, soft tissue deformation can cause substantial guidance errors within these systems. In this dissertation, a mechanics-based approach for modeling soft tissue deformation is proposed to compensate for these effects during image-to-physical registration. A novel method for linearized reconstruction of intraoperative deformation is developed to rapidly estimate volumetric organ deformations from sparse intraoperative organ measurements. Additionally, intraoperative ultrasound is incorporated to extend the fidelity of deformable registration beyond conventional accuracy limits of image-to-physical registration. Finally, a model is created for predicting spatial distributions of deformable registration error from intraoperative patterns of data coverage, through developing a lower bound for registration uncertainty based on the dissipation of elastic energy from data constraints. This model for registration uncertainty can ensure high registration accuracy by directing the locality of data collection, and provide a real-time intraoperative assessment of registration confidence to inform operative risk when using image guidance to navigate and localize. The development of these frameworks leads to superior localization accuracy and quantifiable certainty in the fidelity of image-guided interventions, to improve the precision and safety of therapeutic delivery. | |
