Development of Computational Tools for Computerized Training of Cryosurgery

Cryosurgery is the destruction of undesired tissue by freezing. Modern cryosurgery is performed by strategically inserting a number of cooling probes (cryoprobes) shaped like long hypodermic needles into the target region. Minimally invasive cryosurgery presents unique challenges to the clinician associated with the use of ultrasound (US) as the imaging modality, the selection of the cryoprobe configuration, the accurate placement of the cryoprobes, and the monitoring of the procedure through imaging and temperature measurements. Currently the cost of training clinicians to perform cryosurgery is exceptionally high because the procedure itself is very complex, and the learning process consists of a long residency period. The work presented in this thesis is part of an ongoing project between the Biothermal Technology Laboratory and the Computational Engineering and Robotics Laboratory at Carnegie Mellon University, to develop a computerized training platform that will teach clinicians how to perform the minimally invasive cryosurgery effectively and efficiently. This work uses prostate cryosurgery as a developmental model for verification and benchmarking of all algorithms. Towards that goal, this work provides many of the tools needed to create a virtual training environment. A cryosurgical training framework was presented to enable the diverse functionality of such a computerized training system. Next, a method for generating synthetic ultrasound images from a small library of samples was presented. This method is based off of a texture synthesis technique called “Image Analogies”. The algorithm generates realistic 3D ultrasound images from a small library of samples. Then, an improvement was made to an efficient bioheat transfer simulator. The implementation used GPU computing to achieve a 15x increase in performance over the previous state of the art approach. The enhanced bioheat simulator was leveraged in a cryosurgically relevant ultrasound simulator. The ultrasound simulator uses nonlinear ray-tracing and a novel energy propagation step to simulate imaging artifacts associated with cryoprobe insertion and the thermal field in real-time. Finally, an enhancement to the cryoprobe placement algorithm “bubble packing” was performed to enable the addition of geometric constraints. These geometric constraints allow more clinically relevant plans to be generated. It is the hope of the author that these tools will be the foundation of a fully immersive surgical simulator.