Warning:
This page is
provided for historical and archival purposes
only. While the seminar dates are correct, we offer no
guarantee of informational accuracy or link
validity. Contact information for the speakers, hosts and
seminar committee are certainly out of date.
AI Lab
Computer Science Department
University of Chicago
Roentgen aids in the design of radiation therapy plans for cancer patients using case-based reasoning (CBR). It can function as an aide to the dosimetrist---the person responsible for designing therapy plans---by suggesting first-approximation therapy plans for a new patient, by giving an account of flaws that occur in the simulated results (e.g., under-dosed tumor tissue, over-dosed spinal cord), and by suggesting and applying repairs to a therapy plan to reduce or eliminate flaws. Alternatively, the system can function autonomously to develop finished therapy plans.
Roentgen relies on a memory of good plans developed by dosimetrists for past therapy patients in order to perform its tasks. Given a new patient, Roentgen retrieves the closest match to the patient from this case memory. It tailors the plan to the geometric specifics of the new case to arrive at a first-approximation plan. The system's repair abilities are also supported by a case memory, in this case, a memory of repair episodes previously performed by the dosimetrist to reduce similar flaws in an earlier therapy plan.
Roentgen demonstrates the feasibility of the CBR approach in radiation therapy planning, a problem domain that has resisted efforts based on mathematical optimization and rule-based expert systems. It is an interactive system that fits well with the natural problem solving style of the human user. It extends the applicability of CBR methods by showing how to retrieve cases in a domain where geometric considerations are very important. Finally, it shows the promise of a system building strategy that uses preexisting archives of problem solutions constructed by human experts.