| Surgical procedures, when guided by pre-operative images and tracked tools, can be compromised by the errors involved in registering the images and instruments to the patient. The process of minimizing these errors is analogous to optimizing quality on a manufacturing production line. For example, when building machinery or components to be assembled, the designer must specify a set of tolerances which describe the range in which a specific part can be built and still be properly assembled such that it will perform within the desired specifications. If the tools used cannot build the parts within the specified tolerances then the quality of the end product will suffer. Quality engineers regularly inspect the components being assembled to ensure that they meet the required specifications, so that costs are minimized and quality is maximized.;However, due to the inherent uncertainties in the tracking technology and registration methods, the locations of the surgical targets, planned surgical path and tolerance boundaries are not known exactly, but rather take on a statistical distribution. At specific points of interest (targets), the uncertainties can be described by statistics of the target registration error (TRE), which is a function of the registration algorithms, the specific geometry and the uncertainty of identifying the landmarks, i.e., fiducial localizer error (FLE), used to infer the location of the point of interest. If the TRE statistics can be computed for a locus of points that define the targets, paths and boundaries, then the uncertainties of those features can be shown graphically to the surgeon via a comprehensive surgical plan, so that the surgeon can make strategic decisions pre- and peri-operatively to ensure success.;In this thesis, a complete framework is developed for estimating the overall error for IGI systems that use point-based registration, optical and/or magnetic tracking. In particular, TRE statistical models are developed for point-based registration where the FLE is either (i) anisotropic and homogeneous or (ii) anisotropic and non- homogeneous. An algorithm that estimates the FLE and TRE in real-time is presented, which is useful for optical tracking systems that use point-based registration to determine the tool pose. Furthermore, statistical models are developed for tracking systems based on magnetic fields where there is no explicit registration but the FLE uncertainty models are available for sensors that report tracking information with five and six degrees of freedom. In this case, a new term, target tracking error (TTE), is introduced which is very similar to THE but carries a distinct name to emphasize the lack of a registration taking place. Finally, the application of these models is examined in context of an intracardiac surgical guidance system where tracked 2D ultrasound is a feature.;The work presented here, in concert with related work in the literature, provides a complete framework to estimate the uncertainty associated with the patient to image registration and tracking systems. The error estimates allow for the development of comprehensive surgical plans, improves patient registration algorithms, and provides the surgeon with a powerful tool for quality control during IGI procedures.;A similar framework can be applied to image-guided interventions (IGI), which comprise both surgical (IGS) and therapeutic (IGT) procedures, where the surgeon is generally trying either to (i) resect a section of tissue, (ii) implant a device, (iii) apply a dose of therapeutic energy, or (iv) apply some combination of the three. Before surgery, the surgeon uses the surgical workstation to develop a pre-operative plan. On this plan, the path to the target is identified along with critical areas to avoid such as nerves, blood vessels, etc. A set of boundaries can be drawn to identify the critical areas along the path which essentially act as tolerances, where if the surgical tools move outside these boundaries additional complications may occur. Since IGI systems are based upon tracking technologies and image registration, the tool locations can be monitored to ensure that they fall within the tolerance boundaries.;Keywords: Image-guided surgery, therapy and interventions; image registration; minimally invasive surgery; pre-operative planning; and error modeling and analysis. |
