Stereotaxic radiosurgery is a procedure that uses a narrowly focused beam of radiation as an ablative surgical instrument to destroy brain tumors (more generally, brain lesions). The beam is produced by a linear accelerator that is moved by a mechanical gantry. Radiation is concentrated by crossfiring at the tumor from multiple directions, to reduce the amount of energy deposited in healthy tissues. The treatment planning problem (also called the inverse dosimetry problem) is to find a set of feasible beam configurations that will result in a dose distribution throughout tissues that satisfies input constraints. Because tumors vary in size and shape, as well as in their location relative to critical structures, which should receive very low doses, solving this problem can be very difficult and time-consuming. In most radiosurgical systems, the mechanical gantry has restricted kinematics, which reduces the complexity of the planning problem, but also the quality of the treatment. The Cyberknife is a new system designed and implemented at Stanford Medical School, whose gantry is a general 6-degree-of-freedom robot arms. In this talk I will present the planner that we have developed for this system and I will show experimental results on multiple tumor types, including tumors in the brain, spine, and prostate.

This talk will present material developed by Rhea Tombroupoulos in he PhD thesis. This work was done jointly with John Adler in the Neurosurgery Department.

Enter U of H from Entrance 1 on Calhoun. Go to the visitors booth, tell them your destination. You will be given a map, directions, a parking token, and instructions.

Motion Planning with Visibility Constraints:
Building Autonomous Observers

Autonomous Observers are mobile robots that cooperatively perform vision tasks. Their design raises motion planning problems of a new type, where visibility constraints and motion obstructions must be simultaneously taken into account. We call these problems "motion planning with visibility constraints". I will first introduce the concept of an Autonomous Observer and its applications. I will then present three problems in motion planning with visibility constraints: model building, target finding, and target tracking. Finally, I will develop in more detail the target-finding problem: Given an environment cluttered by view-obstructing obstacles, how many observers are needed to reliably find a moving target hiding in this environment? How to plan a guaranteed motion strategy for these observers? I will show experimental results with implemented planners.

The work presented in this talk was done jointly with Leo Guibas, Hector Gonzalez, Steve LaValle, David Lin, Rajeev Motwani, and Carlo Tomasi.