Computational modeling in support of cancer treatment (Adventures in medical physics and biomedical informatics)

Ira Kalet
UW Radiation Oncology

Over the last 40 years, the subfield of physics known as "medical physics" has evolved and developed in many directions. One area of focus is on mathematical and computational modeling for design of radiation treatment for cancer. Medical physicists have been leaders in building interactive simulation systems for radiation treatment planning. While most have focused on the mathematical methods for computing radiation dose from an arbitrary radiation machine configuration, others have emphasized elaborate computer graphics for user interaction. At the University of Washington we have taken a third direction, the application of artificial intelligence ideas, e.g. rule-based and probabilistic reasoning systems and heuristic search, to solve problems that are not readily modeled with traditional methods. One example is the use of a symbolic topological model of anatomy to generate Markov chain models to predict metastatic spread of tumor cells from the primary tumor to lymph nodes in the surrounding regions. This strategy could provide a significant improvement in accuracy in defining the target of radiation treatment, thus reducing the chance of irradiating tissues not containing tumor and increasing the probability of covering all suspect regions.

The talk will include a short demonstration of the Prism interactive modeling system built at the UW, and used for many years in actual clinical practice.