|
Despite a number of recent advances in robot-assisted surgery, achieving minimal access still requires that surgeons operate with reduced faculties for perception and manipulation as compared to open surgery. Image guidance shows promise for enhancing perception during local navigation (e.g. near occluded endophytic tumors), and we hypothesize that these methods can be extended to address the global navigation problem of efficiently locating and exposing a target organ and its associated anatomical structures. In this work we describe the high-level architecture of an augmented reality system for guiding access to abdominal organs in laparoscopic and robot-assisted procedures, and demonstrate the applicability of an array of assimilation algorithms through proof-of-concept simulation. Under the proposed framework, a coarse model of procedure-specific internal anatomy is initialized based on segmented pre-operative imaging. The model is rigidly registered to the patient at the time of trocar placement, then non-rigidly updated in an incremental manner during the access phase of surgery based on limited views of relevant anatomical structures as they are exposed. Observations are assumed to derive primarily from reconstruction of stereoscopic imaging; however, the assimilation framework provides a means of incorporating measurements made with other sensing modalities. Simulation results show that standard state estimation algorithms are suitable for accommodating large-scale displacement and deformation of the observed feature configuration relative to the initial model. Future work will include development of a suitable 3D model of anatomical structures involved in partial nephrectomy as well as provision for leveraging intraoperative dynamics in the assimilation framework.
|
